QuorumCoverage should only count unknown nodes #2333
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Quorum coverage checks if we have heard from enough nodes to know that there is no entry that can have been written to enough nodes that we haven't heard from to have formed an ack quorum. The coverage algorithm was correct pre-5e399df. 5e399df(BOOKKEEPER-759: Delay Ensemble Change & Disable Ensemble Change) broke this, but it still seems to have worked because they had a broken else statement at the end. Why a change which is 100% about the write-path changed something in the read-path is a mystery. dcdd1e(Small fix wrong nodesUninitialized count when checkCovered) went on to fix the broken fix, so the whole thing ended up broke. The change also modifies ReadLastConfirmedOp to make it testable.
Contributor
Author
|
@sijie @rdhabalia @merlimat you already approved the original patch Please any of you take a look and confirm your approval |
Contributor
Author
|
It looks like I still have a test to fix: |
sijie
approved these changes
May 19, 2020
Contributor
Author
|
@sijie did all CI tests pass this time ? |
Member
|
@eolivelli all CI tests passed. |
Member
|
hi @eolivelli, I have a question that why remove following code snippet? thanks After i learning the mechanism of bookkeeper, i tried to fix this flaky test case.
|
Contributor
Author
|
That test was wrong. |
Member
right, 💯 |
Ghatage
pushed a commit
to Ghatage/bookkeeper
that referenced
this pull request
Jun 19, 2020
The original patch was contributed by ivankelly in PR apache#2303, I have only fixed checkstyle and removed two tests that were wrong. Quorum coverage checks if we have heard from enough nodes to know that there is no entry that can have been written to enough nodes that we haven't heard from to have formed an ack quorum. The coverage algorithm was correct pre-5e399df. 5e399df(BOOKKEEPER-759: Delay Ensemble Change & Disable Ensemble Change) broke this, but it still seems to have worked because they had a broken else statement at the end. Why a change which is 100% about the write-path changed something in the read-path is a mystery. dcdd1e(Small fix wrong nodesUninitialized count when checkCovered) went on to fix the broken fix, so the whole thing ended up broke. The change also modifies ReadLastConfirmedOp to make it testable. Reviewers: Sijie Guo <None>, Rajan Dhabalia <rdhabalia@apache.org> This closes apache#2333 from eolivelli/pr2303
codelipenghui
pushed a commit
to apache/pulsar
that referenced
this pull request
Nov 18, 2020
It would be great to update branch-2.6 to a stable version of bk. so some issues could be fixed. e.g. apache/bookkeeper#2333
Contributor
|
just fyi: fix: #2333 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The original patch was contributed by @ivankelly in PR #2303, I have only fixed checkstyle and removed two tests that were wrong.
Quorum coverage checks if we have heard from enough nodes to know that
there is no entry that can have been written to enough nodes that we
haven't heard from to have formed an ack quorum.
The coverage algorithm was correct pre-5e399df.
5e399df(BOOKKEEPER-759: Delay Ensemble Change & Disable Ensemble
Change) broke this, but it still seems to have worked because they had
a broken else statement at the end. Why a change which is 100% about
the write-path changed something in the read-path is a mystery.
dcdd1e(Small fix wrong nodesUninitialized count when checkCovered)
went on to fix the broken fix, so the whole thing ended up broke.
The change also modifies ReadLastConfirmedOp to make it testable.