>From the output you shared, it looks like you have at least one *very* large partition — my guess is its lots of very small rows per partition. This would explain why the slow compaction and why "nodetool stop compaction" didn’t work (prior to CASSANDRA-14397 stop only works between partitions).
That large partition may cause problems elsewhere if its read. Jordan On Thu, Jul 5, 2018 at 9:35 AM, atul atri <atulatri2...@gmail.com> wrote: > Hi Cris, > > Thank you for reply. > > I already have tried to run "nodetool stop compaction" and this does not > help. I have restarted each node in cluster one by one and compaction > starts again. It gets stuck on same table. > > Following in 'nodetool compactionstats' output. It's stuck at *1336035468* > for more than 35 hours at least. > > > > > *pending tasks: 1 compaction type keyspace > table completed total unit progress > Compactionnotification_system_v1user_notification 1336035468 > 1660997721 bytes 80.44%Active compaction remaining time : 0h00m38s* > > > Following is output for "nodetool cfstats". > > > > > > > > > > > > > > > > > > > > > > > *Table: user_notification SSTable count: 18 Space used > (live), bytes: 17247516201 Space used (total), bytes: 17316488652 > SSTable Compression Ratio: 0.41922805938461566 Number of keys > (estimate): 32556160 Memtable cell count: 44717 Memtable data > size, bytes: 27705294 Memtable switch count: 5 Local read > count: 0 Local read latency: 0.000 ms Local write count: > 236961 Local write latency: 0.047 ms Pending tasks: 0 > Bloom filter false positives: 0 Bloom filter false ratio: 0.00000 > Bloom filter space used, bytes: 72414688 Compacted partition > minimum bytes: 104 Compacted partition maximum bytes: 4966933177 > Compacted partition mean bytes: 1183 Average live cells per > slice (last five minutes): 0.0 Average tombstones per slice (last > five minutes): 0.0* > > Please let me know if any more information. I am really thankful to you > for spending time on this investigation. > > Thanks & Regards, > Atul Atri. > > > On 5 July 2018 at 20:54, Chris Lohfink <clohf...@apple.com> wrote: > >> That looks a bit to me like it isnt stuck but just a long running >> compaction. Can you include the output of `nodetool compactionstats` and >> the `nodetool cfstats` with schema for the table thats being compacted >> (redacted names if necessary). >> >> Can stop compaction with `nodetool stop COMPACTION` or restarting the >> node. >> >> Chris >> >> On Jul 5, 2018, at 12:08 AM, atul atri <atulatri2...@gmail.com> wrote: >> >> Hi, >> >> We noticed that compaction process is also hanging on a node in backup >> ring. Please find attached thread dump for both servers. Recently, we have >> made few changes in cluster topology. >> >> a. Added new server in backup data-center and decommissioned old server. >> Backup ring only has 2 server. >> b. Added new node in primary data-center. Now it has 4 nods. >> >> Is there way we can stop this compaction? As we have added a new node in >> this cluster and we are waiting to run cleanup on this node on which >> compaction is hanging. I am afraid that cleanup will not start until >> compaction job finishes. >> >> Attachments: >> 1. cass-logg02.prod2.thread_dump.out: Thread dump from old node in >> primary datacenter >> 2. cass-logg03.prod1.thread_dump.out: Thread dump from new node in >> backup datacenter. This node is added recently. >> >> Your help is much appreciated. >> >> Thanks & Regards, >> Atul Atri. >> >> >> On 4 July 2018 at 21:15, atul atri <atulatri2...@gmail.com> wrote: >> >>> Hi Chris, >>> Thanks for reply. >>> >>> Unfortunately, our servers do not have jstack installed. >>> I tried "kill -3 <PID>" option but that is also not generating thread >>> dump. >>> >>> Is there any other way I can generate thread dump? >>> >>> Thanks & Regards, >>> Atul Atri. >>> >>> On 4 July 2018 at 20:32, Chris Lohfink <clohf...@apple.com> wrote: >>> >>>> Can you take a thread dump (jstack) and share the state of the >>>> compaction threads? Also check for “Exception” in logs >>>> >>>> Chris >>>> >>>> Sent from my iPhone >>>> >>>> On Jul 4, 2018, at 8:37 AM, atul atri <atulatri2...@gmail.com> wrote: >>>> >>>> Hi, >>>> >>>> On one of our server, compaction process is hanging. It's stuck at 80%. >>>> It was stuck for last 3 days. And today we did a cluster restart (one host >>>> at time). And again it is stuck at same 80%. CPU usages are 100% and there >>>> seems no IO issue. We are seeing following kinds of WARNING in system.log >>>> >>>> *BatchStatement.java (line 226) Batch of prepared statements for [****, >>>> *****] is of size 7557, exceeding specified threshold of 5120 by 2437.* >>>> >>>> >>>> Other than this there seems no error. I have tried to stop compaction >>>> process, but it does not stop. Cassandra version is 2.1. >>>> >>>> Can someone please guide us in solving this issue? >>>> >>>> Thanks & Regards, >>>> Atul Atri. >>>> >>>> >>> >> <cass-logg02.prod2.thread_dump.out><cass-logg03.prod1.thread_dump.out> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org >> <user-unsubscr...@cassandra.apache.org> >> For additional commands, e-mail: user-h...@cassandra.apache.org >> <user-h...@cassandra.apache.org> >> >> >> >
