[v24.1.x] Fixed state_machine_manager stall
#18638
Merged
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
If neither of state machines managed by the state machine manager can accept the foreground apply there is no need to read the range of batches from the log. In this case we can skip the foreground apply read and move on directly to background apply dispatch. Signed-off-by: Michał Maślanka <michal@redpanda.com> (cherry picked from commit 08af2b2)
If an apply results in error the stm manager should yield to prevent busy looping and consuming a lot of CPU cycles. Fixed yielding in background apply fiber and added a yield if foreground apply didn't make progress. Signed-off-by: Michał Maślanka <michal@redpanda.com> (cherry picked from commit e121770)
In state machine manager the background apply fiber is there to apply batches to state machines which are behind the main apply fiber. The background apply fiber is only active if an stm is behind what is currently being applied by the main fiber. A bug in reader configuration lead to a situation in which a background apply could read up to `_next` offset of the state machine manager leading to the situation in which a stm `last_applied_offset` is equal to `_next` therefore the stm next offset is actually equal to `state_machine_manager::_next + 1`. This lead to a situation in which there may be no stms suitable to be handled in main apply fiber but also they are not recoverable by background apply fiber as their `_next` offset is greater of the one from the state machine manager. This causes a stall in the `state_machine_manager` and state machines stop making progress. Signed-off-by: Michał Maślanka <michal@redpanda.com> (cherry picked from commit 0f8963c)
Signed-off-by: Michał Maślanka <michal@redpanda.com> (cherry picked from commit 139823e)
Signed-off-by: Michał Maślanka <michal@redpanda.com> (cherry picked from commit 1abe995)
piyushredpanda
approved these changes
May 23, 2024
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.
Backport of PR #18626