objstorage: read ahead is broken #2531
Comments
RaduBerinde
added a commit
to RaduBerinde/pebble
that referenced
this issue
May 16, 2023
This change improves the objstorage test to allow reading arbitrary parts of objects. We add a test that verifies that the read-ahead implementation does the Prefetch and the reopen that is expected. We also improve logging FS to log `ReadAt` and `Prefetch` calls. Informs cockroachdb#2531.
|
@jbowens @sumeerbhola - I recall we had a perf issue ~recently and we suspected readahead. Could this be related? Or is this only due to recent changes to VFS related to objstorage refactorings, etc. |
The problem exists on crl-release-23.1 too unfortunately. |
|
If we're talking about the same issue, it was on 22.2 and we ended up suspecting that it was an issue of the OS page cache being trashed instead of loss of readahead. |
RaduBerinde
added a commit
to RaduBerinde/pebble
that referenced
this issue
May 16, 2023
This change improves the objstorage test to allow reading arbitrary parts of objects. We add a test that verifies that the read-ahead implementation does the Prefetch and the reopen that is expected. We also improve logging FS to log `ReadAt` and `Prefetch` calls. Informs cockroachdb#2531.
RaduBerinde
added a commit
that referenced
this issue
May 16, 2023
This change improves the objstorage test to allow reading arbitrary parts of objects. We add a test that verifies that the read-ahead implementation does the Prefetch and the reopen that is expected. We also improve logging FS to log `ReadAt` and `Prefetch` calls. Informs #2531.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The
readaheadStatein the vfs read handle is not being initialized properly, resulting in no "automatic" read-ahead functionality (read-ahead for compactions still works).The text was updated successfully, but these errors were encountered: