changefeedccl: fix unwatched column families memory monitoring bug #154802
Conversation
|
This change is |
andyyang890
force-pushed
the
20251003-fix-alloc-leak
branch
from
fb86bc0 to
3804925
Compare
andyyang890
changed the title
andyyang890
added
backport-24.1.x
andyyang890
force-pushed
the
20251003-fix-alloc-leak
branch
3 times, most recently
from
445e667 to
fc8482b
Compare
| } | ||
|
|
||
| // Update watched column again to verify the feed is still progressing. | ||
| // If the memory allocations leaked, this assertion would time out. |
There was a problem hiding this comment.
Is there a way to get this to fail with a clearer symptom than a timeout (at least for enterprise feeds)? This is my first time looking at the memory monitor, but my sense was that since this would set a low PerChangefeedMemLimit, that the memory monitors should fail the feed once the limit is passed. But when I tried to check that on the status of the feed in this test, I saw that it was still running.
There was a problem hiding this comment.
If the allocations leaked, the memory monitor wouldn't fail; instead, the changefeed would get stuck, which is why the assertion would time out. I added some more explanatory comments to the test.
andyyang890
force-pushed
the
20251003-fix-alloc-leak
branch
from
fc8482b to
a14e13a
Compare
This patch fixes a bug where the memory monitor wouldn't reclaim the memory allocated to events corresponding to unwatched column families for a changefeed that targets only a subset of a table's families. Release note (bug fix): A bug where a changefeed targeting only a subset of a table's column families could become stuck has been fixed.
andyyang890
force-pushed
the
20251003-fix-alloc-leak
branch
from
a14e13a to
4f701c2
Compare
|
TFTRs! bors r=aerfrei,asg0451 |
![blathers[crl]](https://avatars.githubusercontent.com/in/59700?s=40&u=95b789b1d594ac6b74f377aa7555d8d98d63240d&v=4){kind=avatar}
|
Encountered an error creating backports. Some common things that can go wrong:
You might need to create your backport manually using the backport tool. 💡 Consider backporting to the fork repo instead of the main repo. See instructions for more details. error creating merge commit from 4f701c2 to blathers/backport-release-24.1-154802: POST https://api.github.com/repos/cockroachdb/cockroach/merges: 409 Merge conflict [] you may need to manually resolve merge conflicts with the backport tool. Backport to branch 24.1.x failed. See errors above. 💡 Consider backporting to the fork repo instead of the main repo. See instructions for more details. error creating merge commit from 4f701c2 to blathers/backport-release-24.3-154802: POST https://api.github.com/repos/cockroachdb/cockroach/merges: 409 Merge conflict [] you may need to manually resolve merge conflicts with the backport tool. Backport to branch 24.3.x failed. See errors above. 💡 Consider backporting to the fork repo instead of the main repo. See instructions for more details. error creating merge commit from 4f701c2 to blathers/backport-release-25.2-154802: POST https://api.github.com/repos/cockroachdb/cockroach/merges: 409 Merge conflict [] you may need to manually resolve merge conflicts with the backport tool. Backport to branch 25.2.x failed. See errors above. 💡 Consider backporting to the fork repo instead of the main repo. See instructions for more details. error creating merge commit from 4f701c2 to blathers/backport-release-25.3-154802: POST https://api.github.com/repos/cockroachdb/cockroach/merges: 409 Merge conflict [] you may need to manually resolve merge conflicts with the backport tool. Backport to branch 25.3.x failed. See errors above. 💡 Consider backporting to the fork repo instead of the main repo. See instructions for more details. 🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf. |
This patch fixes a bug where the memory monitor wouldn't reclaim the
memory allocated to events corresponding to unwatched column families
for a changefeed that targets only a subset of a table's families.
Fixes #154776
Release note (bug fix): A bug where a changefeed targeting only a subset
of a table's column families could become stuck has been fixed.