confirm pivot block - removed disconnect based on chain height estimate #6889

macfarla · 2024-04-05T10:08:35Z

PR description

When choosing the "best" peer to try to confirm the pivot block

~~use the default comparator (TTD then chain height)~~ effectively the same
remove the pre-requirement for the chain height estimate
don't disconnect the "worst" useless peer at this point
and update the peer's chain height estimate if the pivot block is confirmed
also allow 5 peers below the pivot block height since the chain height estimate we use for comparison can be wildly inaccurate

Fixed Issue(s)

Fixes #6887

Thanks for sending a pull request! Have you done the following?

Checked out our contribution guidelines?
Considered documentation and added the doc-change-required label to this PR if updates are required.
Considered the changelog and included an update if required.
For database changes (e.g. KeyValueSegmentIdentifier) considered compatibility and performed forwards and backwards compatibility tests

Locally, you can run these tests to catch failures early:

unit tests: ./gradlew build
acceptance tests: ./gradlew acceptanceTest
integration tests: ./gradlew integrationTest
reference tests: ./gradlew ethereum:referenceTests:referenceTests

Signed-off-by: Sally MacFarlane <[email protected]>

…estimate-tolerance

Signed-off-by: Sally MacFarlane <[email protected]>

…estimate-tolerance

macfarla · 2024-04-07T04:10:21Z

3 EthPeers {
PeerId: 0x3bd1127b27212d2d... 1297018 height (est) PeerReputation score: 97, timeouts: {1=4, 3=1}, useless: 0, validated? true, disconnected? false, client: erigon/v2.58.1-f12e451c/linux-amd64/go1.21.7, [Connection with hashCode 820003208 inboundInitiated? false initAt 1712461468564], enode://3bd1127b27212d2d165ffaa7ae1e0efbdfdf7ff2022e851dc95271e44bb271446ab05e54035756da7eb6ce86b595992d94f2400ed19a34d9485df81fc12d0879@147.28.163.12:30304, 
PeerId: 0xac906289e4b7f12d... 1296969 height (est) PeerReputation score: 150, timeouts: {}, useless: 0, validated? true, disconnected? false, client: Geth/v1.13.14-stable-2bd6bd01/linux-amd64/go1.21.7, [Connection with hashCode 1951103145 inboundInitiated? false initAt 1712461466663], enode://ac906289e4b7f12df423d654c5a962b6ebe5b3a74cc9e06292a85221f9a64a6f1cfdd6b714ed6dacef51578f92b34c60ee91e9ede9c7f8fadc4d347326d95e2b@146.190.13.128:30303, 
PeerId: 0xa3435a0155a3e837... 1297002 height (est) PeerReputation score: 150, timeouts: {}, useless: 0, validated? true, disconnected? false, client: Geth/v1.13.14-stable-2bd6bd01/linux-amd64/go1.21.7, [Connection with hashCode 455188871 inboundInitiated? false initAt 1712461466663], enode://a3435a0155a3e837c02f5e7f5662a2f1fbc25b48e4dc232016e1c51b544cb5b4510ef633ea3278c0e970fa8ad8141e2d4d0f9f95456c537ff05fdf9b31c15072@178.128.136.233:30303}

macfarla · 2024-04-07T04:12:58Z

2024-04-07 13:45:03.763	
trailing peer requirements TrailingPeerRequirements{minimumHeightToBeUpToDate=1296943, maxTrailingPeers=5}

2024-04-07 13:45:03.763	
trailing peer requirements TrailingPeerRequirements{minimumHeightToBeUpToDate=0, maxTrailingPeers=9223372036854775807}

gfukushima · 2024-04-07T23:33:02Z

...eum/eth/src/main/java/org/hyperledger/besu/ethereum/eth/sync/fastsync/SyncTargetManager.java

@@ -74,7 +74,8 @@ public SyncTargetManager(
  protected CompletableFuture<Optional<EthPeer>> selectBestAvailableSyncTarget() {
    final BlockHeader pivotBlockHeader = fastSyncState.getPivotBlockHeader().get();
    final EthPeers ethPeers = ethContext.getEthPeers();
-    final Optional<EthPeer> maybeBestPeer = ethPeers.bestPeerWithHeightEstimate();
+    // just use the default (merge) comparator
+    final Optional<EthPeer> maybeBestPeer = ethPeers.bestPeer();


I think this might be a step back thinking about PoS networks, no much use for the comparison using TTD since this is basically a constant after chain switches to PoS.

I think it actually evaluates to the same thing in the end so I'm reverting this line change.

Signed-off-by: Sally MacFarlane <[email protected]>

…estimate-tolerance

macfarla · 2024-04-09T07:19:19Z

I started 3 holesky nodes last night - 2 of 3 did really well, the third one dropped to zero peers (this seems to happen on holesky, workaround is to add the 2 bootnodes using admin_addPeer so they effectively become static peers). The first 2 still have 25 peers and the third one still has only 2 peers (the bootnodes) <- I have seen this on holesky independent of this PR (maybe worth investigating as a discovery issue separately but I don't think it's high priority)

holesky sync also fine (once the zero peers node got peers, it quickly got back in sync)

macfarla · 2024-04-09T07:20:49Z

2 x mainnet nodes started today, one took a long time (more than 5h) to find peers - mainnet-sally-6889-07 - currently on 24 peers. Both look to be syncing fine (so far)

sync progress

gfukushima · 2024-04-10T06:32:36Z

...um/eth/src/main/java/org/hyperledger/besu/ethereum/eth/sync/fastsync/FastSyncDownloader.java

@@ -163,7 +163,7 @@ public void deleteFastSyncState() {
  protected FastSyncState updateMaxTrailingPeers(final FastSyncState state) {
    if (state.getPivotBlockNumber().isPresent()) {
      trailingPeerRequirements =
-          Optional.of(new TrailingPeerRequirements(state.getPivotBlockNumber().getAsLong(), 0));
+          Optional.of(new TrailingPeerRequirements(state.getPivotBlockNumber().getAsLong(), 5));


My suggestion is, we should probably eliminate this TrailingPeer validation (since there's no garantee that peers are going to have their chain height estimate up-to-date) but for test purpose we can easily skip the trailing validation by assigning Long.MAX_VALUE to the second param of the TrailingPeerRequirements constructor

…ectly disconnecting good peers Signed-off-by: Gabriel Fukushima <[email protected]>

…estimate-tolerance

…/besu into chain-height-estimate-tolerance

macfarla · 2024-06-25T04:26:34Z

this has been incorporated into #6968

macfarla added 14 commits April 5, 2024 05:48

gst

e64112f

Signed-off-by: Sally MacFarlane <[email protected]>

update chain height estimate of that peer to reflect reality

e44fcc2

Signed-off-by: Sally MacFarlane <[email protected]>

log details on comparator update

da0454a

Signed-off-by: Sally MacFarlane <[email protected]>

use default comparator not chain height

9a374ab

Signed-off-by: Sally MacFarlane <[email protected]>

extra logging re best peer

77b9ea7

Signed-off-by: Sally MacFarlane <[email protected]>

allow 1 trailing peer not zero

d01a423

Signed-off-by: Sally MacFarlane <[email protected]>

removed check for chain height prior to confirming pivot block

500843d

Signed-off-by: Sally MacFarlane <[email protected]>

remove a log line that is repeated unnecessarily

c825c47

Signed-off-by: Sally MacFarlane <[email protected]>

cleanup

5ebcdac

Signed-off-by: Sally MacFarlane <[email protected]>

Merge branch 'main' of github.com:hyperledger/besu into chain-height-…

8f5ac92

…estimate-tolerance

removed unused method

7db08e9

Signed-off-by: Sally MacFarlane <[email protected]>

allow 5 trailing peers not 0

ea60f72

Signed-off-by: Sally MacFarlane <[email protected]>

add chain height estimate to ethPeer.toString()

d3d15f6

Signed-off-by: Sally MacFarlane <[email protected]>

Merge branch 'main' of github.com:hyperledger/besu into chain-height-…

75da10f

…estimate-tolerance

gfukushima reviewed Apr 7, 2024

View reviewed changes

logging order

efe7a64

Signed-off-by: Sally MacFarlane <[email protected]>

macfarla changed the title ~~confirm pivot block - removed requirement for chain height estimate~~ confirm pivot block - removed disconnect based on chain height estimate Apr 8, 2024

macfarla added 4 commits April 8, 2024 13:08

revert change in comparator

755d41b

Signed-off-by: Sally MacFarlane <[email protected]>

update test

2ea6a32

Signed-off-by: Sally MacFarlane <[email protected]>

merge

a3dceb4

Signed-off-by: Sally MacFarlane <[email protected]>

Merge branch 'main' of github.com:hyperledger/besu into chain-height-…

20dbba0

…estimate-tolerance

macfarla marked this pull request as draft April 9, 2024 11:46

gfukushima reviewed Apr 10, 2024

View reviewed changes

gfukushima and others added 3 commits April 10, 2024 16:43

Merge branch 'main' into chain-height-estimate-tolerance

2d81237

use Long.MAX_VALUE to skip the trailingPeer validation, likely incorr…

8352324

…ectly disconnecting good peers Signed-off-by: Gabriel Fukushima <[email protected]>

Merge branch 'main' of github.com:hyperledger/besu into chain-height-…

47fae0a

…estimate-tolerance

Merge branch 'chain-height-estimate-tolerance' of github.com:macfarla…

94f3a3e

…/besu into chain-height-estimate-tolerance

macfarla closed this Jun 25, 2024

macfarla deleted the chain-height-estimate-tolerance branch December 5, 2024 00:13

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

confirm pivot block - removed disconnect based on chain height estimate #6889

confirm pivot block - removed disconnect based on chain height estimate #6889

macfarla commented Apr 5, 2024 •

edited

Loading

macfarla commented Apr 7, 2024

macfarla commented Apr 7, 2024

gfukushima Apr 7, 2024

macfarla Apr 8, 2024

macfarla commented Apr 9, 2024

macfarla commented Apr 9, 2024 •

edited

Loading

gfukushima Apr 10, 2024 •

edited

Loading

macfarla commented Jun 25, 2024

confirm pivot block - removed disconnect based on chain height estimate #6889

confirm pivot block - removed disconnect based on chain height estimate #6889

Conversation

macfarla commented Apr 5, 2024 • edited Loading

PR description

Fixed Issue(s)

Thanks for sending a pull request! Have you done the following?

Locally, you can run these tests to catch failures early:

macfarla commented Apr 7, 2024

macfarla commented Apr 7, 2024

gfukushima Apr 7, 2024

Choose a reason for hiding this comment

macfarla Apr 8, 2024

Choose a reason for hiding this comment

macfarla commented Apr 9, 2024

macfarla commented Apr 9, 2024 • edited Loading

gfukushima Apr 10, 2024 • edited Loading

Choose a reason for hiding this comment

macfarla commented Jun 25, 2024

macfarla commented Apr 5, 2024 •

edited

Loading

macfarla commented Apr 9, 2024 •

edited

Loading

gfukushima Apr 10, 2024 •

edited

Loading