Repository navigation
[Store] Disarm the persistent write-failure injector before the reuse retry in dfs_sync_client_test - #4785
Closed
he-yufeng wants to merge 1 commit into
Closed
[Store] Disarm the persistent write-failure injector before the reuse retry in dfs_sync_client_test#4785he-yufeng wants to merge 1 commit into
he-yufeng wants to merge 1 commit into
Conversation
… retry in dfs_sync_client_test DfsOnlyBatchFailureOnlyRevokesFailedKey arms FailingPosixFsAdapter with a persistent fault (the design defeats the backend's reopen-retry so the batch leg surfaces the error), then expects the revoked key's reuse Put to succeed. With the injector still armed, every later WriteAt keeps failing, so the retry can never succeed: the test failed deterministically on main CI (both the post-merge run and every PR's CTest since kvcache-ai#4637). Disarm the injector before the retry leg, which is exactly the scenario the second half models: the transient fault has cleared once the failed key has been revoked. Verified in a Linux container: the test fails at the reuse assertion before this change and passes after it; the full dfs_sync_client_test suite goes 16 passed / 2 environment-skipped.
he-yufeng
requested review from
XucSh,
YiXR,
stmatengss and
ykwd
as code owners
October 19, 2026 22:11
10 of 36 tasks
Collaborator
Author
|
Closing as a duplicate of the earlier #4781 — same root cause (the persistent injector fails the reuse retry), same disarm mechanism, verified equivalent in a Linux container (target test passes, suite 16 passed / 2 environment-skipped). Left a small review note there about keeping the persistent-failure design comment. Whichever lands first unblocks the CI Gate on #4764. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary
DfsOnlyBatchFailureOnlyRevokesFailedKey(added in #4637) has been failing CTest deterministically on main and on every PR's CI since (e.g. main run 39242136165, and #4764's gate).The test arms
FailingPosixFsAdapterwith a persistent write fault — the comment in the adapter says this is deliberate, so the backend's one-shot reopen-retry cannot mask the batch-leg failure. But the second half of the test then expects the revoked key's reusePutto succeed, which is impossible while the injector is still armed: every laterWriteAtfails. The reuse assertion at line 635 fails deterministically (Actual: false).Fix: disarm the injector (
FailWriteCall(-1)) before the retry leg. That is exactly the scenario the second half models — the transient fault has cleared once the failed key has been revoked.Verification
Linux container (ubuntu 22.04, posix adapter): before the change the test fails at the reuse assertion; after, it passes. Full
dfs_sync_client_testsuite: 16 passed, 2 environment-skipped (the checksum cases skip the same way they do on CI).