HI Asad, The post flush task frees up allocated commit log segments.
Apart for commit log segment allocation the post flush task "synchronises custom secondary indexes and provides ordering guarantees for futures on switchMemtable/flush etc, which expect to be able to wait until the flush (and all prior flushes) requested have completed." The post flush executor is a single threaded thread pool that cannot be tuned (https://github.com/apache/cassandra/blob/8b3a60b9a7dbefeecc06bace617279612ec7092d/src/java/org/apache/cassandra/db/ColumnFamilyStore.java#L160-L167 <https://github.com/apache/cassandra/blob/8b3a60b9a7dbefeecc06bace617279612ec7092d/src/java/org/apache/cassandra/db/ColumnFamilyStore.java#L160-L167>) Are you using secondary indexes? I there a high write throughput which is resulting in frequent meltable flushes? Regards, Akhil > On 12/08/2017, at 7:05 AM, ZAIDI, ASAD A <az1...@att.com> wrote: > > Hello Folks, > > I’m using Cassandra 2.2 on 14 node cluster. > > Now a days, I’m observing memtablepostflush pending number going high , this > happens intermittently. I’m looking if Is there way to ‘tune’ > memtablepostflush stage? > > Thanks/ASad
