[
https://issues.apache.org/jira/browse/IMPALA-12721?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17815208#comment-17815208
]
ASF subversion and git services commented on IMPALA-12721:
----------------------------------------------------------
Commit 99e8170997f18db0f63d451af89ca32320ebb465 in impala's branch
refs/heads/master from Yida Wu
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=99e817099 ]
IMPALA-12721: Fix flaky tests involving check_deleted_file_fd()
check_deleted_file_fd() is introduced in IMPALA-12681, however some
spilling testcases involving check_deleted_file_fd() seem flaky.
This patch fixed the issue by adding a retry mechanism within the
check_deleted_file_fd() function. If the function encounters a
failure, it retries the process of verifying the presence of a
deleted referencing file. Based on my local test, the file will be
removed after the test even when the test fails and the call to
delete the file handle is ahead of the call to remove the file (This
has been confirmed through additional testing logs). While there is
no theory why this would happen, introducing a retry mechanism has
allowed the test case to run successfully for 200 times without
encountering any failures. It is possible that a delay may be
occurring at some point in the process which leads to this kind of
failure.
Tests:
Reran the testcase 200 times without a failure.
Change-Id: I900aab7dc9833015ce140253ff40da28a6ed3ba6
Reviewed-on: http://gerrit.cloudera.org:8080/21000
Reviewed-by: Impala Public Jenkins <[email protected]>
Tested-by: Impala Public Jenkins <[email protected]>
> TestScratchDir.test_scratch_dirs_remote_spill seems flaky
> ---------------------------------------------------------
>
> Key: IMPALA-12721
> URL: https://issues.apache.org/jira/browse/IMPALA-12721
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Reporter: Yida Wu
> Assignee: Yida Wu
> Priority: Major
>
> In exhaustive release builds, sometimes the testcase fails with following
> messages, which is related to the patch of IMPALA-12681:
> {code:java}
> custom_cluster/test_scratch_disk.py:100: in teardown_method
> self.check_deleted_file_fd() custom_cluster/test_scratch_disk.py:305: in
> check_deleted_file_fd assert deleted_files is None E assert '46283626 0
> lr-x------ 1 jenkins jenkins 64 Jan 16 14:29 /proc/21537/fd/440 ->
> /tmp/tmpZwObRe/impala-scra...29 /proc/21537/fd/453 ->
> /tmp/tmpZwObRe/impala-scratch/impala-scratch-bdeb9c30-b2d9-4355-9070-646176770c1e\\
> (deleted)' is None
> {code}
> However, in a local environment, it was observed that the files in the fd
> path could be deleted after the test failure. Therefore, it might be
> necessary to update the testcase to prevent false-positive checks.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
