[
https://issues.apache.org/jira/browse/KAFKA-9450?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17024902#comment-17024902
]
Navinder Brar commented on KAFKA-9450:
--------------------------------------
Or do you suggest to never (ie, for EOS and non-EOS case) call
`innerByteStore#flush()`? This might be possible but would have a negative
impact on non-EOS as it would make current fault-tolerance mechanism for
non-EOS less efficient (we would not have a guarantee on commit that data is
flushed to disk and might need to recover more data from the changelog topic in
case of failure).
>>> [~mjsax] do you mean we still write to checkpoint file(for non-EOS) on
>>> every commit but remove flush. That would be dangerous right? As if the
>>> data from Rockdsdb is not flushed for the checkpoints that have been
>>> written in checkpoint file we have lost the data and moved ahead as well.
Can we add event listeners on Rocksdb(EventListener::OnFlushCompleted()) and
whenever a particular store is flushed, commit the checkpoint for that
particular store(changelog) in the checkpoint file. Currently, we are
overriding most performance-based Rocksdb configs(memtable size, max writer
buffers) by making the commit based on time. If this seems reasonable, I can
work on this.
> Decouple inner state flushing from committing with EOS
> ------------------------------------------------------
>
> Key: KAFKA-9450
> URL: https://issues.apache.org/jira/browse/KAFKA-9450
> Project: Kafka
> Issue Type: Improvement
> Components: streams
> Reporter: Sophie Blee-Goldman
> Priority: Major
>
> When EOS is turned on, the commit interval is set quite low (100ms) and all
> the store layers are flushed during a commit. This is necessary for
> forwarding records in the cache to the changelog, but unfortunately also
> forces rocksdb to flush the current memtable before it's full. The result is
> a large number of small writes to disk, losing the benefits of batching, and
> a large number of very small L0 files that are likely to slow compaction.
> Since we have to delete the stores to recreate from scratch anyways during an
> unclean shutdown with EOS, we may as well skip flushing the innermost
> StateStore during a commit and only do so during a graceful shutdown, before
> a rebalance, etc. This is currently blocked on a refactoring of the state
> store layers to allow decoupling the flush of the caching layer from the
> actual state store.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
