tests: flush in test_isolation #10658
Conversation
| @@ -104,7 +104,12 @@ def post_checks(env: NeonEnv, test_output_dir: Path, db_name: str, endpoint: End | |||
| if env.pg_version <= PgVersion.V15 and r[0] == "i": | |||
| unlogged_relation_files.append(f"{r[1]}_init") | |||
|
|
|||
| ignored_files = unlogged_relation_files | |||
| ignored_files = unlogged_relation_files + [ | |||
There was a problem hiding this comment.
@hlinnaka could you please think about whether it's expected that we see these files differ (by being absent) when doing a check_restored_datadir_content after test_isolation? These differences don't exist after a normal run of pg_regress, but I presume that's because they're just not being "touched" in test_isolation?
(I only have a vague grasp of what's going on inside these postgres suites)
7414 tests run: 7061 passed, 0 failed, 353 skipped (full report)Flaky tests (2)Postgres 17
Postgres 14
Code coverage* (full report)
* collected from Rust tests only The comment gets automatically updated with the latest test results
eff5aff at 2025-02-05T17:56:08.753Z :recycle: |
jcsp
force-pushed
the
jcsp/issue-10654-tests
branch
from
a523187 to
f21fc65
Compare
jcsp
force-pushed
the
jcsp/issue-10654-tests
branch
from
f21fc65 to
5407174
Compare
jcsp
force-pushed
the
jcsp/issue-10654-tests
branch
from
5407174 to
eff5aff
Compare
| filename.startswith("pg_xact") | ||
| or filename.startswith("pg_multixact") |
There was a problem hiding this comment.
Hmm. If a pg_xact or pg_multixact file is downloaded on-demand, then it might be missing from the running postgres cluster, but present in the pageserver. But how can it really happen the other way 'round? How can you have a file in the running postgres instance that's not present in the pageserver?
Problem
This test occasionally fails while the test teardown tries to do a graceful shutdown, because the test has quickly written lots of data into the pageserver.
Closes: #10654
Summary of changes
post_checksat the end oftest_isolation, as we already do for test_pg_regress -- this improves our detection of issues, and as a nice side effect flushes the pageserver.check_restored_datadir_content. I am optimistically presuming that these differences (present in test_isolation) are okay, as as far as we know there's nothing critically broken here, but will get expert review.