PG-1932 Make sure archive and restore comamnds do not leak files #467
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
jeltz
force-pushed
the
archive-restore-file-leak
branch
2 times, most recently
from
7b2fcfb to
c6a89ac
Compare
jeltz
changed the title
Codecov Report❌ Patch coverage is ❌ Your project status has failed because the head coverage (59.66%) is below the target coverage (90.00%). You can increase the head coverage or adjust the target coverage. Additional details and impacted files@@ Coverage Diff @@
## main #467 +/- ##
==========================================
- Coverage 59.69% 59.66% -0.03%
==========================================
Files 67 67
Lines 10471 10481 +10
Branches 1813 1809 -4
==========================================
+ Hits 6251 6254 +3
- Misses 3524 3532 +8
+ Partials 696 695 -1
🚀 New features to boost your workflow:
|
jeltz
force-pushed
the
archive-restore-file-leak
branch
from
c6a89ac to
ff8e70b
Compare
dAdAbird
approved these changes
Nov 13, 2025
Member
dAdAbird
left a comment
dAdAbird
left a comment
There was a problem hiding this comment.
LGTM overall. The only thing, maybe it worth putting a comment about that race somewhere in the code? So it'd be less questions and investigation when someone looks at the code later on.
As reported by Roman Leventov in percona/postgres#582 we had an issue where we easily could leak temporary files and directories when on failure, especially when the wrapped restore or archive command fails. The solution in this commit is to move the cleanup of temporary files and directories to and atexit() callback plus add a signal handler for SIGINT and SIGTERM. The signal handler makes sure to kill itself after running cleanup so the parent process gets the right error. We have a tiny race condition between creating the temporary directory and registering the signal handler but that does not seem worth fixing since it would just leak a directory, not the WAL file, the race conditions being tiny. With the current hardcoded path of /dev/shm and the TAP test framework testing this seems a pain so no tests are included.
jeltz
force-pushed
the
archive-restore-file-leak
branch
from
ff8e70b to
78700ca
Compare
Collaborator
Author
|
Thanks, added comments. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
As reported by Roman Leventov in percona/postgres#582 we had an issue where we easily could leak temporary files and directories when on failure, especially when the wrapped restore or archive command fails.
The solution in this commit is to move the cleanup of temporary files and directories to an
atexit()callback plus add a signal handler forSIGINTandSIGTERM. The signal handler makes sure to kill itself after running cleanup so the parent process gets the right error.We have a tiny race condition between creating the temporary directory and registering the signal handler but that does not seem worth fixing since it would just leak a directory, not the WAL file, the race conditions being tiny.
With the current hardcoded path of
/dev/shmand the TAP test framework testing this seems a pain so no tests are included.Question: Should I add TAP tests or is it ok that I skipped them?