diegomrsantos commented on code in PR #4092:
URL: https://github.com/apache/iggy/pull/4092#discussion_r3982697140
##########
core/shard/src/lib.rs:
##########
@@ -7381,6 +7416,24 @@ where
// spreads over every group instead of replaying the same prefix.
rotate_sweep_to_cursor(namespace_scratch,
self.partition_walk_cursor.get());
+ let mut persistence_metrics =
partitions::PersistenceMetrics::default();
+ for namespace in namespace_scratch.iter() {
Review Comment:
I’ve added a [proposed
fix](https://github.com/diegomrsantos/iggy/commit/f48a26aca19cef48d4858973ac31417801892dc8)
alongside the
[regression](https://github.com/diegomrsantos/iggy/commit/76248cf515e9523c4e746bba9d9ded627e306c80),
both on top of `7343be420`.
The fix withholds acknowledgments at the shared partition send function and
stops persistence processing after fencing. Checkpoint preparation and queued
request resumption now return fatal failures directly to the tick.
The regression now passes on Linux and macOS. Additional coverage includes a
real checkpoint index write failure while the WAL remains healthy, plus healthy
primary and backup acknowledgment paths. The partition and shard unit tests
pass on both platforms.
The two commits are intended to be reviewed and applied together, since the
regression fails on its own. Would you prefer to cherry-pick the pair, or would
a small PR targeting `topic_durability` be more convenient? Happy to adjust the
approach.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
