WAL redo processes leak DSM segments, i.e. files in /dev/shm/
#9738
Assignees
Labels
c/compute
Component: compute, excluding postgres itself
c/storage/pageserver
Component: storage: pageserver
t/bug
Issue Type: Bug
triaged
bugs that were already triaged
Comments
hlinnaka
added
c/compute
|
After this is fixed, rolled out, and won't be rolled back, ping so we can do the cleanup. |
knizhnik
added a commit
that referenced
this issue
Nov 19, 2024
## Problem See #9738 ## Summary of changes Do not create DSM segment for wal_redo Postgres --------- Co-authored-by: Konstantin Knizhnik <[email protected]>
|
PR solving this problem is merged |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Each WAL redo process creates a tiny DSM segment, like
/dev/shm/PostgreSQL.3449905360. They are not always (never?) cleaned up on exit, not even on pageserver restart, so a pageserver host can accumulate tens of thousands of these files over time. Each file is small, 3-5 kB, but they add up to a lot of wasted memory over time.slack discussion: https://neondb.slack.com/archives/C033RQ5SPDH/p1731486990179549?thread_ts=1731445405.348699&cid=C033RQ5SPDH
A WAL redo process should not need DSM for anything, so let's remove the DSM initialization step altogether, so that the files are not created in the first place.
The text was updated successfully, but these errors were encountered: