Skip to content

fix(test): extend the replication wait time to prevent the flaky test#3531

Merged
PragmaTwice merged 1 commit into
apache:unstablefrom
git-hulk:fix/replication-resume-broken-transfer-flake
Jun 18, 2026
Merged

fix(test): extend the replication wait time to prevent the flaky test#3531
PragmaTwice merged 1 commit into
apache:unstablefrom
git-hulk:fix/replication-resume-broken-transfer-flake

Conversation

@git-hulk

@git-hulk git-hulk commented Jun 18, 2026

Copy link
Copy Markdown
Member

Currently, WaitForSync only allows 5s for master_link_status to reach "up",
which is not enough under loaded CI runners (e.g. SonarCloud coverage builds);

Reproduced locally by running 16 background CPU burners and looping the test:

for i in $(seq 1 6); do
   go test -count=1 -run TestReplicationWithLimitSpeed$ \
      ./integration/replication/... \
      -binPath=.../build/kvrocks -workspace=/tmp/kvrocks-test-ws
done

Before the fix: 1/6 runs failed with the same trace as the CI flake
("Condition never satisfied" at client.go:50, called from replication_test.go:362).
After the fix: 8/8 runs pass under the same CPU load (run times 34-51s vs. an
unloaded baseline of ~28s, confirming load was active).

Related flaky run: https://github.com/apache/kvrocks/actions/runs/27660686644/job/81804342125

🤖 Generated with Claude Code

…sfer flake

Currently, WaitForSync only allows 5s for master_link_status to reach "up",
which is not enough under loaded CI runners (e.g. SonarCloud coverage builds);

Reproduced locally by running 16 background CPU burners and looping the test:

  for i in $(seq 1 6); do
    go test -count=1 -run TestReplicationWithLimitSpeed$ \
        ./integration/replication/... \
        -binPath=.../build/kvrocks -workspace=/tmp/kvrocks-test-ws
  done

Before the fix: 1/6 runs failed with the same trace as the CI flake
("Condition never satisfied" at client.go:50, called from replication_test.go:362).
After the fix: 8/8 runs pass under the same CPU load (run times 34-51s vs. an
unloaded baseline of ~28s, confirming load was active).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@git-hulk git-hulk changed the title fix(test): extend slave master_link_status wait in resume-broken-transfer flake fix(test): extend slave master_link_status wait time to prevent the flaky test Jun 18, 2026
@git-hulk git-hulk changed the title fix(test): extend slave master_link_status wait time to prevent the flaky test fix(test): extend the replication wait time to prevent the flaky test Jun 18, 2026
@jihuayu

jihuayu commented Jun 18, 2026

Copy link
Copy Markdown
Member

This PR is related to the issue #3515.

@PragmaTwice PragmaTwice enabled auto-merge (squash) June 18, 2026 10:08
@PragmaTwice PragmaTwice merged commit d51384c into apache:unstable Jun 18, 2026
75 of 77 checks passed
@sonarqubecloud

Copy link
Copy Markdown

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants