[v25.2.x] storage: pass abort_source to log::start()
#27069
Merged
Conversation
This file contains hidden or 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
`_log->start()` is called from `consensus::do_start()`. Within `log::start()` is a call to `offset_translator().sync_with_log()`, which also passes an `abort_source`. This synchronization function may be a somewhat heavy operation, as a `log_reader` over the range `(_highest_known_offset, dirty_offset]` is used to read the `log`. Previously, the `abort_source` passed to `sync_with_log()` was the `disk_log_impl::_compaction_as`. However, this isn't very helpful because this `abort_source` only has `request_abort()` called upon it within `disk_log_impl::remove()` and `disk_log_impl::close()`- both of which are called at the _end_ of the typical high-level `partition` removal process starting from the `partition_manager`. What this can lead to is the following: 1. `redpanda` is launched. A relatively heavy `partition` with a lagging `_highest_known_offset` enters the `sync_with_log()` operation. 2. The user hits ^C to kill `redpanda`. 3. The high level shutdown process is entered from `application.cc`. 4. Within `partition_manager::do_shutdown(partition)`, we first wait for the `_raft_manager` to shutdown the `partition->raft()` pointer before proceeding to stop the `partition` and then its underlying `log`. 5. In reality, the `raft` pointer hasn't even finished its starting process yet, because the `sync_with_log()` process is taking so long! 6. The user waits for `sync_with_log()` to complete, and only then does shutdown of the `raft` object finish (as the `_compaction_as` only requests abort when the `log` is closed). 7. `log` is closed, `_compaction_as.request_abort()` is called. Clearly, passing `_compaction_as` to `sync_with_log()` is essentially a no-op, since it is guaranteed that `sync_with_log()` _has_ to complete before an abort of `_compaction_as` can be requested. To make the shutdown process smoother and avoid this sub-optimal behavior, pass the `consensus::_as` as an argument to `_log->start()` instead. Then, when an abort is requested of `redpanda` early on, the `sync_with_log()` process will be correctly terminated early instead of potentially appearing as a long winded hang. (cherry picked from commit 9a5c819)
WillemKauf
approved these changes
Jul 30, 2025
CI test resultstest results on build#69971
|
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 #27050