map: fix reconciliation failure caused by out of sync errors number #26742
+119
−11
Conversation
This file contains 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
giorio94
added
kind/bug
|
/test |
giorio94
added
needs-backport/1.14
and removed
backport/1.14
|
@giorio94 Thanks for the patch! Would you be able to craft a regression test for this? This code may undergo significant changes throughout 1.15/1.16. |
Cached maps come with a controller that retries Update/Delete operations in case an error occurred while synchronizing the given entry with the kernel. Currently, it relies on the `outstandingErrors` counter to determine whether reconciliation is necessary, as well as if any entry still needs to be processed in a subsequent operation. Yet, it is possible that this counter gets out of sync, in particular in case the given error is resolved automatically when a subsequent operation acting on the same key succeed. If this happens, the reconciliation function will never complete successfully, as there will continue to be an error that cannot be resolved (as it no longer exists). The given controller will this continue failing forever until the agent gets restarted. Let's fix this inferring the number of outstanding errors from the cache itself. The `outstandingErrors` variable is preserved (although converted to a boolean) to avoid iterating over the full cache in case it is known that no error occurred. Still, if this flag gets out of sync, the only consequence will be that the cache is iterated once to determine that there's actually no failure, ensuring that the reconciliation logic still converges properly. Signed-off-by: Marco Iorio <marco.iorio@isovalent.com>
giorio94
force-pushed
the
mio/bpf-map-sync-resolve-errors
branch
from
8fe6a4c to
ed7ce3a
Compare
I've added a unit test to assert that the error resolver works properly (and succeeds) both in case an element needs to be synchronized, and in case the issue already resolved automatically. The second test consistently fails without the patch introduced in this commit. @ti-mo PTAL. |
|
/test |
1 similar comment
|
/test |
|
Hey @ti-mo, would you have time to take another look at this? |
ti-mo
approved these changes
Aug 16, 2023
maintainer-s-little-helper
bot
added
the
ready-to-merge
tklauser
added
backport-pending/1.14
joestringer
added
backport-done/1.14
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Cached maps come with a controller that retries Update/Delete operations in case an error occurred while synchronizing the given entry with the kernel. Currently, it relies on the
outstandingErrorscounter to determine whether reconciliation is necessary, as well as if any entry still needs to be processed in a subsequent operation.Yet, it is possible that this counter gets out of sync, in particular in case the given error is resolved automatically when a subsequent operation acting on the same key succeed. If this happens, the reconciliation function will never complete successfully, as there will continue to be an error that cannot be resolved (as it no longer exists). The given controller will this continue failing forever until the agent gets restarted.
Let's fix this inferring the number of outstanding errors from the cache itself. The
outstandingErrorsvariable is preserved (although converted to a boolean) to avoid iterating over the full cache in case it is known that no error occurred. Still, if this flag gets out of sync, the only consequence will be that the cache is iterated once to determine that there's actually no failure, ensuring that the reconciliation logic still converges properly.