smiklosovic commented on code in PR #4521:
URL: https://github.com/apache/cassandra/pull/4521#discussion_r2619556710
##########
src/java/org/apache/cassandra/db/compression/CompressionDictionaryCache.java:
##########
@@ -102,7 +102,12 @@ public void add(@Nullable CompressionDictionary
compressionDictionary)
// Only update cache if not already in the cache
DictId newDictId = compressionDictionary.dictId();
- cache.get(newDictId, id -> compressionDictionary);
+ CompressionDictionary cached = cache.get(newDictId, id ->
compressionDictionary);
+ if (cached != compressionDictionary)
+ {
+ compressionDictionary.close();
+ return;
+ }
Review Comment:
this is exercised when doing `refreshDictionaryFromSystemTable` repeatedly
(every `compression_dictionary_refresh_interval`) it is just that in tests,
this will be run also after `compression_dictionary_refresh_initial_delay`
which is 10s so if tests take longer than this then
`refreshDictionaryFromSystemTable` will kick in and it gets here. Same error
would be logged (as in ticket comments) if we just waited one more hour.
My suspicion is that we retrieve the last dict from disk, then we go to
cache it but it is already there so we dont do that, then the rest of the
`cache.add` method finishes and we end up with `CompressionDictionary` from
disk nobody is interested in anymore (because it is cached under a different
instance already) and it is not closed either. So once it is going to be GCed
it will leak - saying that nobody released it - that's because nobody called
`close()` on it.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
