[ https://issues.apache.org/jira/browse/CASSANDRA-15306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17050724#comment-17050724 ]
David Capwell commented on CASSANDRA-15306: ------------------------------------------- Given the fact this is compaction this wold most likely be org.apache.cassandra.cache.ChunkCache I think. org.apache.cassandra.utils.memory.BufferPool.LocalPool#get will attempt to grow until this limit is reached and then will avoid using the pool and rely on GC to cleanup buffers, currently buffers are heap but CASSANDRA-15358 wants to switch to direct (needed by netty). To replicate you can set file_cache_size_in_mb to a very small value, say 0 or 1 (never tried with such a value so may see other issues); this pool is shared between internode networking and check cache > Investigate why we are allocating 8MiB chunks and reaching the maximum > BufferPool size > -------------------------------------------------------------------------------------- > > Key: CASSANDRA-15306 > URL: https://issues.apache.org/jira/browse/CASSANDRA-15306 > Project: Cassandra > Issue Type: Bug > Components: Observability/Logging, Test/benchmark > Reporter: Joey Lynch > Priority: Normal > Fix For: 4.0-beta > > > While throwing some light traffic at {{4.0-alpha1}} I saw a lot of the > following in the logs > {noformat} > INFO [CompactionExecutor:8] 2019-09-06 11:40:31,419 NoSpamLogger.java:91 - > Maximum memory usage reached (512.000MiB), cannot allocate chunk of 8.000MiB > INFO [CompactionExecutor:8] 2019-09-06 11:55:31,419 NoSpamLogger.java:91 - > Maximum memory usage reached (512.000MiB), cannot allocate chunk of 8.000MiB > INFO [CompactionExecutor:15] 2019-09-06 12:10:31,419 NoSpamLogger.java:91 - > Maximum memory usage reached (512.000MiB), cannot allocate chunk of 8.000MiB > INFO [CompactionExecutor:18] 2019-09-06 12:25:31,421 NoSpamLogger.java:91 - > Maximum memory usage reached (512.000MiB), cannot allocate chunk of 8.000MiB > {noformat} > This was with about 150 WPS against a LCS table containing 4kib partitions. > It seemed that compaction proceeded just fine but I don't remember seeing > this in previous testing runs and I'd like to make sure it's not a bug > (otherwise we may want to reduce the logging). -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org For additional commands, e-mail: commits-h...@cassandra.apache.org