2015-07-07 03:54:22 +03:00
|
|
|
// Copyright 2015 The go-ethereum Authors
|
2015-07-22 19:48:40 +03:00
|
|
|
// This file is part of the go-ethereum library.
|
2015-07-07 03:54:22 +03:00
|
|
|
//
|
2015-07-23 19:35:11 +03:00
|
|
|
// The go-ethereum library is free software: you can redistribute it and/or modify
|
2015-07-07 03:54:22 +03:00
|
|
|
// it under the terms of the GNU Lesser General Public License as published by
|
|
|
|
// the Free Software Foundation, either version 3 of the License, or
|
|
|
|
// (at your option) any later version.
|
|
|
|
//
|
2015-07-22 19:48:40 +03:00
|
|
|
// The go-ethereum library is distributed in the hope that it will be useful,
|
2015-07-07 03:54:22 +03:00
|
|
|
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
2015-07-22 19:48:40 +03:00
|
|
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
2015-07-07 03:54:22 +03:00
|
|
|
// GNU Lesser General Public License for more details.
|
|
|
|
//
|
|
|
|
// You should have received a copy of the GNU Lesser General Public License
|
2015-07-22 19:48:40 +03:00
|
|
|
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
2015-07-07 03:54:22 +03:00
|
|
|
|
2015-01-27 15:33:26 +02:00
|
|
|
package discover
|
|
|
|
|
|
|
|
import (
|
|
|
|
"crypto/ecdsa"
|
|
|
|
"fmt"
|
|
|
|
"math/rand"
|
|
|
|
"net"
|
|
|
|
"reflect"
|
2024-05-29 16:02:26 +03:00
|
|
|
"slices"
|
2015-01-27 15:33:26 +02:00
|
|
|
"testing"
|
|
|
|
"testing/quick"
|
2015-06-10 16:15:35 +03:00
|
|
|
"time"
|
2015-01-27 15:33:26 +02:00
|
|
|
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
"github.com/ethereum/go-ethereum/common/mclock"
|
2015-01-27 15:33:26 +02:00
|
|
|
"github.com/ethereum/go-ethereum/crypto"
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
"github.com/ethereum/go-ethereum/internal/testlog"
|
|
|
|
"github.com/ethereum/go-ethereum/log"
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
"github.com/ethereum/go-ethereum/p2p/enode"
|
|
|
|
"github.com/ethereum/go-ethereum/p2p/enr"
|
2019-01-31 13:48:54 +03:00
|
|
|
"github.com/ethereum/go-ethereum/p2p/netutil"
|
2015-01-27 15:33:26 +02:00
|
|
|
)
|
|
|
|
|
2015-03-25 17:45:53 +02:00
|
|
|
func TestTable_pingReplace(t *testing.T) {
|
2018-02-12 15:36:09 +03:00
|
|
|
run := func(newNodeResponding, lastInBucketResponding bool) {
|
|
|
|
name := fmt.Sprintf("newNodeResponding=%t/lastInBucketResponding=%t", newNodeResponding, lastInBucketResponding)
|
|
|
|
t.Run(name, func(t *testing.T) {
|
|
|
|
t.Parallel()
|
|
|
|
testPingReplace(t, newNodeResponding, lastInBucketResponding)
|
|
|
|
})
|
|
|
|
}
|
2015-04-27 01:50:18 +03:00
|
|
|
|
2018-02-12 15:36:09 +03:00
|
|
|
run(true, true)
|
|
|
|
run(false, true)
|
|
|
|
run(true, false)
|
|
|
|
run(false, false)
|
|
|
|
}
|
2015-03-25 17:45:53 +02:00
|
|
|
|
2018-02-12 15:36:09 +03:00
|
|
|
func testPingReplace(t *testing.T, newNodeIsResponding, lastInBucketIsResponding bool) {
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
simclock := new(mclock.Simulated)
|
2018-02-12 15:36:09 +03:00
|
|
|
transport := newPingRecorder()
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
tab, db := newTestTable(transport, Config{
|
|
|
|
Clock: simclock,
|
|
|
|
Log: testlog.Logger(t, log.LevelTrace),
|
|
|
|
})
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
defer db.Close()
|
2019-04-30 14:13:22 +03:00
|
|
|
defer tab.close()
|
2015-03-25 17:45:53 +02:00
|
|
|
|
2018-02-12 15:36:09 +03:00
|
|
|
<-tab.initDone
|
2015-01-27 15:33:26 +02:00
|
|
|
|
2018-07-03 16:24:12 +03:00
|
|
|
// Fill up the sender's bucket.
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
replacementNodeKey, _ := crypto.HexToECDSA("45a915e4d060149eb4365960e6a7a45f334393093061116b197e3240065ff2d8")
|
2024-05-29 16:02:26 +03:00
|
|
|
replacementNode := enode.NewV4(&replacementNodeKey.PublicKey, net.IP{127, 0, 0, 1}, 99, 99)
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
last := fillBucket(tab, replacementNode.ID())
|
|
|
|
tab.mutex.Lock()
|
|
|
|
nodeEvents := newNodeEventRecorder(128)
|
|
|
|
tab.nodeAddedHook = nodeEvents.nodeAdded
|
|
|
|
tab.nodeRemovedHook = nodeEvents.nodeRemoved
|
|
|
|
tab.mutex.Unlock()
|
2015-01-27 15:33:26 +02:00
|
|
|
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
// The revalidation process should replace
|
|
|
|
// this node in the bucket if it is unresponsive.
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
transport.dead[last.ID()] = !lastInBucketIsResponding
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
transport.dead[replacementNode.ID()] = !newNodeIsResponding
|
|
|
|
|
|
|
|
// Add replacement node to table.
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(replacementNode, false)
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
|
|
|
|
t.Log("last:", last.ID())
|
|
|
|
t.Log("replacement:", replacementNode.ID())
|
|
|
|
|
|
|
|
// Wait until the last node was pinged.
|
|
|
|
waitForRevalidationPing(t, transport, tab, last.ID())
|
|
|
|
|
|
|
|
if !lastInBucketIsResponding {
|
|
|
|
if !nodeEvents.waitNodeAbsent(last.ID(), 2*time.Second) {
|
|
|
|
t.Error("last node was not removed")
|
|
|
|
}
|
|
|
|
if !nodeEvents.waitNodePresent(replacementNode.ID(), 2*time.Second) {
|
|
|
|
t.Error("replacement node was not added")
|
|
|
|
}
|
|
|
|
|
|
|
|
// If a replacement is expected, we also need to wait until the replacement node
|
|
|
|
// was pinged and added/removed.
|
|
|
|
waitForRevalidationPing(t, transport, tab, replacementNode.ID())
|
|
|
|
if !newNodeIsResponding {
|
|
|
|
if !nodeEvents.waitNodeAbsent(replacementNode.ID(), 2*time.Second) {
|
|
|
|
t.Error("replacement node was not removed")
|
|
|
|
}
|
|
|
|
}
|
2015-01-27 15:33:26 +02:00
|
|
|
}
|
|
|
|
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
// Check bucket content.
|
2018-02-12 15:36:09 +03:00
|
|
|
tab.mutex.Lock()
|
|
|
|
defer tab.mutex.Unlock()
|
|
|
|
wantSize := bucketSize
|
|
|
|
if !lastInBucketIsResponding && !newNodeIsResponding {
|
|
|
|
wantSize--
|
|
|
|
}
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
bucket := tab.bucket(replacementNode.ID())
|
|
|
|
if l := len(bucket.entries); l != wantSize {
|
|
|
|
t.Errorf("wrong bucket size after revalidation: got %d, want %d", l, wantSize)
|
2018-02-12 15:36:09 +03:00
|
|
|
}
|
2024-05-29 16:02:26 +03:00
|
|
|
if ok := containsID(bucket.entries, last.ID()); ok != lastInBucketIsResponding {
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
t.Errorf("revalidated node found: %t, want: %t", ok, lastInBucketIsResponding)
|
2018-02-12 15:36:09 +03:00
|
|
|
}
|
|
|
|
wantNewEntry := newNodeIsResponding && !lastInBucketIsResponding
|
2024-05-29 16:02:26 +03:00
|
|
|
if ok := containsID(bucket.entries, replacementNode.ID()); ok != wantNewEntry {
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
t.Errorf("replacement node found: %t, want: %t", ok, wantNewEntry)
|
2018-02-12 15:36:09 +03:00
|
|
|
}
|
2015-01-27 15:33:26 +02:00
|
|
|
}
|
|
|
|
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
// waitForRevalidationPing waits until a PING message is sent to a node with the given id.
|
|
|
|
func waitForRevalidationPing(t *testing.T, transport *pingRecorder, tab *Table, id enode.ID) *enode.Node {
|
|
|
|
t.Helper()
|
2019-01-31 13:48:54 +03:00
|
|
|
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
simclock := tab.cfg.Clock.(*mclock.Simulated)
|
|
|
|
maxAttempts := tab.len() * 8
|
|
|
|
for i := 0; i < maxAttempts; i++ {
|
2024-05-28 22:30:17 +03:00
|
|
|
simclock.Run(tab.cfg.PingInterval * slowRevalidationFactor)
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
p := transport.waitPing(2 * time.Second)
|
|
|
|
if p == nil {
|
|
|
|
t.Fatal("Table did not send revalidation ping")
|
|
|
|
}
|
|
|
|
if id == (enode.ID{}) || p.ID() == id {
|
|
|
|
return p
|
2015-03-30 18:23:28 +03:00
|
|
|
}
|
|
|
|
}
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
t.Fatalf("Table did not ping node %v (%d attempts)", id, maxAttempts)
|
|
|
|
return nil
|
2015-03-30 18:23:28 +03:00
|
|
|
}
|
|
|
|
|
2018-02-12 15:36:09 +03:00
|
|
|
// This checks that the table-wide IP limit is applied correctly.
|
|
|
|
func TestTable_IPLimit(t *testing.T) {
|
|
|
|
transport := newPingRecorder()
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
tab, db := newTestTable(transport, Config{})
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
defer db.Close()
|
2019-04-30 14:13:22 +03:00
|
|
|
defer tab.close()
|
2018-02-12 15:36:09 +03:00
|
|
|
|
|
|
|
for i := 0; i < tableIPLimit+1; i++ {
|
2018-10-12 12:47:24 +03:00
|
|
|
n := nodeAtDistance(tab.self().ID(), i, net.IP{172, 0, 1, byte(i)})
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(n, false)
|
2018-02-12 15:36:09 +03:00
|
|
|
}
|
|
|
|
if tab.len() > tableIPLimit {
|
|
|
|
t.Errorf("too many nodes in table")
|
|
|
|
}
|
2019-01-31 13:48:54 +03:00
|
|
|
checkIPLimitInvariant(t, tab)
|
2018-02-12 15:36:09 +03:00
|
|
|
}
|
|
|
|
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
// This checks that the per-bucket IP limit is applied correctly.
|
2018-02-12 15:36:09 +03:00
|
|
|
func TestTable_BucketIPLimit(t *testing.T) {
|
|
|
|
transport := newPingRecorder()
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
tab, db := newTestTable(transport, Config{})
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
defer db.Close()
|
2019-04-30 14:13:22 +03:00
|
|
|
defer tab.close()
|
2018-02-12 15:36:09 +03:00
|
|
|
|
|
|
|
d := 3
|
|
|
|
for i := 0; i < bucketIPLimit+1; i++ {
|
2018-10-12 12:47:24 +03:00
|
|
|
n := nodeAtDistance(tab.self().ID(), d, net.IP{172, 0, 1, byte(i)})
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(n, false)
|
2018-02-12 15:36:09 +03:00
|
|
|
}
|
|
|
|
if tab.len() > bucketIPLimit {
|
|
|
|
t.Errorf("too many nodes in table")
|
|
|
|
}
|
2019-01-31 13:48:54 +03:00
|
|
|
checkIPLimitInvariant(t, tab)
|
|
|
|
}
|
|
|
|
|
|
|
|
// checkIPLimitInvariant checks that ip limit sets contain an entry for every
|
|
|
|
// node in the table and no extra entries.
|
|
|
|
func checkIPLimitInvariant(t *testing.T, tab *Table) {
|
|
|
|
t.Helper()
|
|
|
|
|
|
|
|
tabset := netutil.DistinctNetSet{Subnet: tableSubnet, Limit: tableIPLimit}
|
|
|
|
for _, b := range tab.buckets {
|
|
|
|
for _, n := range b.entries {
|
|
|
|
tabset.Add(n.IP())
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if tabset.String() != tab.ips.String() {
|
|
|
|
t.Errorf("table IP set is incorrect:\nhave: %v\nwant: %v", tab.ips, tabset)
|
|
|
|
}
|
2018-02-12 15:36:09 +03:00
|
|
|
}
|
|
|
|
|
2020-08-24 15:42:39 +03:00
|
|
|
func TestTable_findnodeByID(t *testing.T) {
|
2015-01-27 15:33:26 +02:00
|
|
|
t.Parallel()
|
|
|
|
|
|
|
|
test := func(test *closeTest) bool {
|
|
|
|
// for any node table, Target and N
|
2018-02-12 15:36:09 +03:00
|
|
|
transport := newPingRecorder()
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
tab, db := newTestTable(transport, Config{})
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
defer db.Close()
|
2019-04-30 14:13:22 +03:00
|
|
|
defer tab.close()
|
2023-12-18 12:47:21 +03:00
|
|
|
fillTable(tab, test.All, true)
|
2015-01-27 15:33:26 +02:00
|
|
|
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
// check that closest(Target, N) returns nodes
|
2020-08-24 15:42:39 +03:00
|
|
|
result := tab.findnodeByID(test.Target, test.N, false).entries
|
2015-01-27 15:33:26 +02:00
|
|
|
if hasDuplicates(result) {
|
|
|
|
t.Errorf("result contains duplicates")
|
|
|
|
return false
|
|
|
|
}
|
|
|
|
if !sortedByDistanceTo(test.Target, result) {
|
|
|
|
t.Errorf("result is not sorted by distance to target")
|
|
|
|
return false
|
|
|
|
}
|
|
|
|
|
|
|
|
// check that the number of results is min(N, tablen)
|
|
|
|
wantN := test.N
|
|
|
|
if tlen := tab.len(); tlen < test.N {
|
|
|
|
wantN = tlen
|
|
|
|
}
|
|
|
|
if len(result) != wantN {
|
|
|
|
t.Errorf("wrong number of nodes: got %d, want %d", len(result), wantN)
|
|
|
|
return false
|
|
|
|
} else if len(result) == 0 {
|
|
|
|
return true // no need to check distance
|
|
|
|
}
|
|
|
|
|
|
|
|
// check that the result nodes have minimum distance to target.
|
|
|
|
for _, b := range tab.buckets {
|
|
|
|
for _, n := range b.entries {
|
2024-05-29 16:02:26 +03:00
|
|
|
if containsID(result, n.ID()) {
|
2015-01-27 15:33:26 +02:00
|
|
|
continue // don't run the check below for nodes in result
|
|
|
|
}
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
farthestResult := result[len(result)-1].ID()
|
|
|
|
if enode.DistCmp(test.Target, n.ID(), farthestResult) < 0 {
|
2015-01-27 15:33:26 +02:00
|
|
|
t.Errorf("table contains node that is closer to target but it's not in result")
|
|
|
|
t.Logf(" Target: %v", test.Target)
|
|
|
|
t.Logf(" Farthest Result: %v", farthestResult)
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
t.Logf(" ID: %v", n.ID())
|
2015-01-27 15:33:26 +02:00
|
|
|
return false
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return true
|
|
|
|
}
|
2015-06-10 16:15:35 +03:00
|
|
|
if err := quick.Check(test, quickcfg()); err != nil {
|
2015-01-27 15:33:26 +02:00
|
|
|
t.Error(err)
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
type closeTest struct {
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
Self enode.ID
|
|
|
|
Target enode.ID
|
2024-05-29 16:02:26 +03:00
|
|
|
All []*enode.Node
|
2015-01-27 15:33:26 +02:00
|
|
|
N int
|
|
|
|
}
|
|
|
|
|
|
|
|
func (*closeTest) Generate(rand *rand.Rand, size int) reflect.Value {
|
|
|
|
t := &closeTest{
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
Self: gen(enode.ID{}, rand).(enode.ID),
|
|
|
|
Target: gen(enode.ID{}, rand).(enode.ID),
|
2015-01-27 15:33:26 +02:00
|
|
|
N: rand.Intn(bucketSize),
|
|
|
|
}
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
for _, id := range gen([]enode.ID{}, rand).([]enode.ID) {
|
2019-01-29 19:39:20 +03:00
|
|
|
r := new(enr.Record)
|
|
|
|
r.Set(enr.IP(genIP(rand)))
|
2024-05-29 16:02:26 +03:00
|
|
|
n := enode.SignNull(r, id)
|
2019-01-29 19:39:20 +03:00
|
|
|
t.All = append(t.All, n)
|
2015-01-27 15:33:26 +02:00
|
|
|
}
|
|
|
|
return reflect.ValueOf(t)
|
|
|
|
}
|
|
|
|
|
2024-05-28 22:30:17 +03:00
|
|
|
func TestTable_addInboundNode(t *testing.T) {
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
tab, db := newTestTable(newPingRecorder(), Config{})
|
2019-01-31 13:48:54 +03:00
|
|
|
<-tab.initDone
|
|
|
|
defer db.Close()
|
2019-04-30 14:13:22 +03:00
|
|
|
defer tab.close()
|
2019-01-31 13:48:54 +03:00
|
|
|
|
|
|
|
// Insert two nodes.
|
|
|
|
n1 := nodeAtDistance(tab.self().ID(), 256, net.IP{88, 77, 66, 1})
|
|
|
|
n2 := nodeAtDistance(tab.self().ID(), 256, net.IP{88, 77, 66, 2})
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(n1, false)
|
|
|
|
tab.addFoundNode(n2, false)
|
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1, n2})
|
2019-01-31 13:48:54 +03:00
|
|
|
|
2024-05-28 22:30:17 +03:00
|
|
|
// Add a changed version of n2. The bucket should be updated.
|
2019-01-31 13:48:54 +03:00
|
|
|
newrec := n2.Record()
|
|
|
|
newrec.Set(enr.IP{99, 99, 99, 99})
|
2024-05-28 22:30:17 +03:00
|
|
|
n2v2 := enode.SignNull(newrec, n2.ID())
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addInboundNode(n2v2)
|
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1, n2v2})
|
2024-05-28 22:30:17 +03:00
|
|
|
|
|
|
|
// Try updating n2 without sequence number change. The update is accepted
|
|
|
|
// because it's inbound.
|
|
|
|
newrec = n2.Record()
|
|
|
|
newrec.Set(enr.IP{100, 100, 100, 100})
|
|
|
|
newrec.SetSeq(n2.Seq())
|
|
|
|
n2v3 := enode.SignNull(newrec, n2.ID())
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addInboundNode(n2v3)
|
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1, n2v3})
|
2019-01-31 13:48:54 +03:00
|
|
|
}
|
|
|
|
|
2024-05-28 22:30:17 +03:00
|
|
|
func TestTable_addFoundNode(t *testing.T) {
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
tab, db := newTestTable(newPingRecorder(), Config{})
|
2019-01-31 13:48:54 +03:00
|
|
|
<-tab.initDone
|
|
|
|
defer db.Close()
|
2019-04-30 14:13:22 +03:00
|
|
|
defer tab.close()
|
2019-01-31 13:48:54 +03:00
|
|
|
|
|
|
|
// Insert two nodes.
|
|
|
|
n1 := nodeAtDistance(tab.self().ID(), 256, net.IP{88, 77, 66, 1})
|
|
|
|
n2 := nodeAtDistance(tab.self().ID(), 256, net.IP{88, 77, 66, 2})
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(n1, false)
|
|
|
|
tab.addFoundNode(n2, false)
|
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1, n2})
|
2019-01-31 13:48:54 +03:00
|
|
|
|
2024-05-28 22:30:17 +03:00
|
|
|
// Add a changed version of n2. The bucket should be updated.
|
2019-01-31 13:48:54 +03:00
|
|
|
newrec := n2.Record()
|
|
|
|
newrec.Set(enr.IP{99, 99, 99, 99})
|
2024-05-28 22:30:17 +03:00
|
|
|
n2v2 := enode.SignNull(newrec, n2.ID())
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(n2v2, false)
|
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1, n2v2})
|
2024-05-28 22:30:17 +03:00
|
|
|
|
|
|
|
// Try updating n2 without a sequence number change.
|
|
|
|
// The update should not be accepted.
|
|
|
|
newrec = n2.Record()
|
|
|
|
newrec.Set(enr.IP{100, 100, 100, 100})
|
|
|
|
newrec.SetSeq(n2.Seq())
|
|
|
|
n2v3 := enode.SignNull(newrec, n2.ID())
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(n2v3, false)
|
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1, n2v2})
|
2024-05-28 22:30:17 +03:00
|
|
|
}
|
2019-01-31 13:48:54 +03:00
|
|
|
|
2024-05-28 22:30:17 +03:00
|
|
|
// This test checks that discv4 nodes can update their own endpoint via PING.
|
|
|
|
func TestTable_addInboundNodeUpdateV4Accept(t *testing.T) {
|
|
|
|
tab, db := newTestTable(newPingRecorder(), Config{})
|
|
|
|
<-tab.initDone
|
|
|
|
defer db.Close()
|
|
|
|
defer tab.close()
|
|
|
|
|
|
|
|
// Add a v4 node.
|
|
|
|
key, _ := crypto.HexToECDSA("dd3757a8075e88d0f2b1431e7d3c5b1562e1c0aab9643707e8cbfcc8dae5cfe3")
|
|
|
|
n1 := enode.NewV4(&key.PublicKey, net.IP{88, 77, 66, 1}, 9000, 9000)
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addInboundNode(n1)
|
2024-05-28 22:30:17 +03:00
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1})
|
|
|
|
|
|
|
|
// Add an updated version with changed IP.
|
|
|
|
// The update will be accepted because it is inbound.
|
|
|
|
n1v2 := enode.NewV4(&key.PublicKey, net.IP{99, 99, 99, 99}, 9000, 9000)
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addInboundNode(n1v2)
|
2024-05-28 22:30:17 +03:00
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1v2})
|
|
|
|
}
|
|
|
|
|
|
|
|
// This test checks that discv4 node entries will NOT be updated when a
|
|
|
|
// changed record is found.
|
|
|
|
func TestTable_addFoundNodeV4UpdateReject(t *testing.T) {
|
|
|
|
tab, db := newTestTable(newPingRecorder(), Config{})
|
|
|
|
<-tab.initDone
|
|
|
|
defer db.Close()
|
|
|
|
defer tab.close()
|
|
|
|
|
|
|
|
// Add a v4 node.
|
|
|
|
key, _ := crypto.HexToECDSA("dd3757a8075e88d0f2b1431e7d3c5b1562e1c0aab9643707e8cbfcc8dae5cfe3")
|
|
|
|
n1 := enode.NewV4(&key.PublicKey, net.IP{88, 77, 66, 1}, 9000, 9000)
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(n1, false)
|
2024-05-28 22:30:17 +03:00
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1})
|
|
|
|
|
|
|
|
// Add an updated version with changed IP.
|
|
|
|
// The update won't be accepted because it isn't inbound.
|
|
|
|
n1v2 := enode.NewV4(&key.PublicKey, net.IP{99, 99, 99, 99}, 9000, 9000)
|
2024-05-29 16:02:26 +03:00
|
|
|
tab.addFoundNode(n1v2, false)
|
2024-05-28 22:30:17 +03:00
|
|
|
checkBucketContent(t, tab, []*enode.Node{n1})
|
|
|
|
}
|
|
|
|
|
|
|
|
func checkBucketContent(t *testing.T, tab *Table, nodes []*enode.Node) {
|
|
|
|
t.Helper()
|
|
|
|
|
|
|
|
b := tab.bucket(nodes[0].ID())
|
|
|
|
if reflect.DeepEqual(unwrapNodes(b.entries), nodes) {
|
|
|
|
return
|
2019-01-31 13:48:54 +03:00
|
|
|
}
|
2024-05-28 22:30:17 +03:00
|
|
|
t.Log("wrong bucket content. have nodes:")
|
|
|
|
for _, n := range b.entries {
|
|
|
|
t.Logf(" %v (seq=%v, ip=%v)", n.ID(), n.Seq(), n.IP())
|
|
|
|
}
|
|
|
|
t.Log("want nodes:")
|
|
|
|
for _, n := range nodes {
|
|
|
|
t.Logf(" %v (seq=%v, ip=%v)", n.ID(), n.Seq(), n.IP())
|
|
|
|
}
|
|
|
|
t.FailNow()
|
|
|
|
|
|
|
|
// Also check IP limits.
|
2019-01-31 13:48:54 +03:00
|
|
|
checkIPLimitInvariant(t, tab)
|
|
|
|
}
|
|
|
|
|
2019-05-15 07:47:45 +03:00
|
|
|
// This test checks that ENR updates happen during revalidation. If a node in the table
|
|
|
|
// announces a new sequence number, the new record should be pulled.
|
|
|
|
func TestTable_revalidateSyncRecord(t *testing.T) {
|
|
|
|
transport := newPingRecorder()
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
tab, db := newTestTable(transport, Config{
|
|
|
|
Clock: new(mclock.Simulated),
|
|
|
|
Log: testlog.Logger(t, log.LevelTrace),
|
|
|
|
})
|
2019-05-15 07:47:45 +03:00
|
|
|
<-tab.initDone
|
|
|
|
defer db.Close()
|
|
|
|
defer tab.close()
|
|
|
|
|
|
|
|
// Insert a node.
|
|
|
|
var r enr.Record
|
|
|
|
r.Set(enr.IP(net.IP{127, 0, 0, 1}))
|
|
|
|
id := enode.ID{1}
|
2024-05-29 16:02:26 +03:00
|
|
|
n1 := enode.SignNull(&r, id)
|
|
|
|
tab.addFoundNode(n1, false)
|
2019-05-15 07:47:45 +03:00
|
|
|
|
|
|
|
// Update the node record.
|
|
|
|
r.Set(enr.WithEntry("foo", "bar"))
|
|
|
|
n2 := enode.SignNull(&r, id)
|
|
|
|
transport.updateRecord(n2)
|
|
|
|
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
// Wait for revalidation. We wait for the node to be revalidated two times
|
2024-05-24 12:33:19 +03:00
|
|
|
// in order to synchronize with the update in the table.
|
p2p/discover: improved node revalidation (#29572)
Node discovery periodically revalidates the nodes in its table by sending PING, checking
if they are still alive. I recently noticed some issues with the implementation of this
process, which can cause strange results such as nodes dropping unexpectedly, certain
nodes not getting revalidated often enough, and bad results being returned to incoming
FINDNODE queries.
In this change, the revalidation process is improved with the following logic:
- We maintain two 'revalidation lists' containing the table nodes, named 'fast' and 'slow'.
- The process chooses random nodes from each list on a randomized interval, the interval being
faster for the 'fast' list, and performs revalidation for the chosen node.
- Whenever a node is newly inserted into the table, it goes into the 'fast' list.
Once validation passes, it transfers to the 'slow' list. If a request fails, or the
node changes endpoint, it transfers back into 'fast'.
- livenessChecks is incremented by one for successful checks. Unlike the old implementation,
we will not drop the node on the first failing check. We instead quickly decay the
livenessChecks give it another chance.
- Order of nodes in bucket doesn't matter anymore.
I am also adding a debug API endpoint to dump the node table content.
Co-authored-by: Martin HS <martin@swende.se>
2024-05-23 15:26:09 +03:00
|
|
|
waitForRevalidationPing(t, transport, tab, n2.ID())
|
|
|
|
waitForRevalidationPing(t, transport, tab, n2.ID())
|
|
|
|
|
2019-05-15 07:47:45 +03:00
|
|
|
intable := tab.getNode(id)
|
|
|
|
if !reflect.DeepEqual(intable, n2) {
|
|
|
|
t.Fatalf("table contains old record with seq %d, want seq %d", intable.Seq(), n2.Seq())
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2022-12-08 01:31:47 +03:00
|
|
|
func TestNodesPush(t *testing.T) {
|
|
|
|
var target enode.ID
|
|
|
|
n1 := nodeAtDistance(target, 255, intIP(1))
|
|
|
|
n2 := nodeAtDistance(target, 254, intIP(2))
|
|
|
|
n3 := nodeAtDistance(target, 253, intIP(3))
|
2024-05-29 16:02:26 +03:00
|
|
|
perm := [][]*enode.Node{
|
2022-12-08 01:31:47 +03:00
|
|
|
{n3, n2, n1},
|
|
|
|
{n3, n1, n2},
|
|
|
|
{n2, n3, n1},
|
|
|
|
{n2, n1, n3},
|
|
|
|
{n1, n3, n2},
|
|
|
|
{n1, n2, n3},
|
|
|
|
}
|
|
|
|
|
|
|
|
// Insert all permutations into lists with size limit 3.
|
|
|
|
for _, nodes := range perm {
|
|
|
|
list := nodesByDistance{target: target}
|
|
|
|
for _, n := range nodes {
|
|
|
|
list.push(n, 3)
|
|
|
|
}
|
2024-05-29 16:02:26 +03:00
|
|
|
if !slices.EqualFunc(list.entries, perm[0], nodeIDEqual) {
|
2022-12-08 01:31:47 +03:00
|
|
|
t.Fatal("not equal")
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Insert all permutations into lists with size limit 2.
|
|
|
|
for _, nodes := range perm {
|
|
|
|
list := nodesByDistance{target: target}
|
|
|
|
for _, n := range nodes {
|
|
|
|
list.push(n, 2)
|
|
|
|
}
|
2024-05-29 16:02:26 +03:00
|
|
|
if !slices.EqualFunc(list.entries, perm[0][:2], nodeIDEqual) {
|
2022-12-08 01:31:47 +03:00
|
|
|
t.Fatal("not equal")
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2024-05-29 16:02:26 +03:00
|
|
|
func nodeIDEqual[N nodeType](n1, n2 N) bool {
|
2022-12-08 01:31:47 +03:00
|
|
|
return n1.ID() == n2.ID()
|
|
|
|
}
|
|
|
|
|
2015-01-27 15:33:26 +02:00
|
|
|
// gen wraps quick.Value so it's easier to use.
|
|
|
|
// it generates a random value of the given value's type.
|
|
|
|
func gen(typ interface{}, rand *rand.Rand) interface{} {
|
|
|
|
v, ok := quick.Value(reflect.TypeOf(typ), rand)
|
|
|
|
if !ok {
|
|
|
|
panic(fmt.Sprintf("couldn't generate random value of type %T", typ))
|
|
|
|
}
|
|
|
|
return v.Interface()
|
|
|
|
}
|
|
|
|
|
2019-01-29 19:39:20 +03:00
|
|
|
func genIP(rand *rand.Rand) net.IP {
|
|
|
|
ip := make(net.IP, 4)
|
|
|
|
rand.Read(ip)
|
|
|
|
return ip
|
|
|
|
}
|
|
|
|
|
all: new p2p node representation (#17643)
Package p2p/enode provides a generalized representation of p2p nodes
which can contain arbitrary information in key/value pairs. It is also
the new home for the node database. The "v4" identity scheme is also
moved here from p2p/enr to remove the dependency on Ethereum crypto from
that package.
Record signature handling is changed significantly. The identity scheme
registry is removed and acceptable schemes must be passed to any method
that needs identity. This means records must now be validated explicitly
after decoding.
The enode API is designed to make signature handling easy and safe: most
APIs around the codebase work with enode.Node, which is a wrapper around
a valid record. Going from enr.Record to enode.Node requires a valid
signature.
* p2p/discover: port to p2p/enode
This ports the discovery code to the new node representation in
p2p/enode. The wire protocol is unchanged, this can be considered a
refactoring change. The Kademlia table can now deal with nodes using an
arbitrary identity scheme. This requires a few incompatible API changes:
- Table.Lookup is not available anymore. It used to take a public key
as argument because v4 protocol requires one. Its replacement is
LookupRandom.
- Table.Resolve takes *enode.Node instead of NodeID. This is also for
v4 protocol compatibility because nodes cannot be looked up by ID
alone.
- Types Node and NodeID are gone. Further commits in the series will be
fixes all over the the codebase to deal with those removals.
* p2p: port to p2p/enode and discovery changes
This adapts package p2p to the changes in p2p/discover. All uses of
discover.Node and discover.NodeID are replaced by their equivalents from
p2p/enode.
New API is added to retrieve the enode.Node instance of a peer. The
behavior of Server.Self with discovery disabled is improved. It now
tries much harder to report a working IP address, falling back to
127.0.0.1 if no suitable address can be determined through other means.
These changes were needed for tests of other packages later in the
series.
* p2p/simulations, p2p/testing: port to p2p/enode
No surprises here, mostly replacements of discover.Node, discover.NodeID
with their new equivalents. The 'interesting' API changes are:
- testing.ProtocolSession tracks complete nodes, not just their IDs.
- adapters.NodeConfig has a new method to create a complete node.
These changes were needed to make swarm tests work.
Note that the NodeID change makes the code incompatible with old
simulation snapshots.
* whisper/whisperv5, whisper/whisperv6: port to p2p/enode
This port was easy because whisper uses []byte for node IDs and
URL strings in the API.
* eth: port to p2p/enode
Again, easy to port because eth uses strings for node IDs and doesn't
care about node information in any way.
* les: port to p2p/enode
Apart from replacing discover.NodeID with enode.ID, most changes are in
the server pool code. It now deals with complete nodes instead
of (Pubkey, IP, Port) triples. The database format is unchanged for now,
but we should probably change it to use the node database later.
* node: port to p2p/enode
This change simply replaces discover.Node and discover.NodeID with their
new equivalents.
* swarm/network: port to p2p/enode
Swarm has its own node address representation, BzzAddr, containing both
an overlay address (the hash of a secp256k1 public key) and an underlay
address (enode:// URL).
There are no changes to the BzzAddr format in this commit, but certain
operations such as creating a BzzAddr from a node ID are now impossible
because node IDs aren't public keys anymore.
Most swarm-related changes in the series remove uses of
NewAddrFromNodeID, replacing it with NewAddr which takes a complete node
as argument. ToOverlayAddr is removed because we can just use the node
ID directly.
2018-09-25 01:59:00 +03:00
|
|
|
func quickcfg() *quick.Config {
|
|
|
|
return &quick.Config{
|
|
|
|
MaxCount: 5000,
|
|
|
|
Rand: rand.New(rand.NewSource(time.Now().Unix())),
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-01-27 15:33:26 +02:00
|
|
|
func newkey() *ecdsa.PrivateKey {
|
|
|
|
key, err := crypto.GenerateKey()
|
|
|
|
if err != nil {
|
|
|
|
panic("couldn't generate key: " + err.Error())
|
|
|
|
}
|
|
|
|
return key
|
|
|
|
}
|