[ https://issues.apache.org/jira/browse/CASSANDRA-12215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15386589#comment-15386589 ]
Evan Prothro edited comment on CASSANDRA-12215 at 7/20/16 8:50 PM: ------------------------------------------------------------------- Hau's comment seems related in our minds, since similarly, if we copy to / drop table and delete the data directory from the filesystem / recreate table / copy from and then set gc_grace_seconds to something small, running compaction results in a crash again. Stranger still, if we copy to / drop table and delete the data directory from the filesystem / create a different keyspace / create a differently named table with same schema (with static column) / copy from, and then set gc_grace_seconds to ensure a compaction, running compaction still results in a crash. However, when we copy to / drop table and delete the data directory from the filesystem / recreate table with different schema (same but with no static column) / copy from and then set gc_grace_seconds to something small, we see no tombstones, and compaction runs successfully. We have various other anecdotal experiments, but hesitate to expound just yet for fear of confusion. Please let us know if there's something further we can look at or result we can provide to help isolate this. was (Author: eprothro): Hau's comment seems related in our minds, since similarly, if we copy to / clear / recreate table / copy from and then set gc_grace_seconds to something small, running compaction results in a crash again. Stranger still, if we copy to / clear / create a different keyspace / create a differently named table with same schema (with static column) / copy from, and then set gc_grace_seconds to ensure a compaction, running compaction still results in a crash. However, when we copy to / clear / recreate table with different schema (same but with no static column) / copy from and then set gc_grace_seconds to something small, we see no tombstones, and compaction runs successfully. We have various other anecdotal experiments, but hesitate to expound just yet for fear of confusion. Please let us know if there's something further we can look at or result we can provide to help isolate this. > NullPointerException during Compaction > -------------------------------------- > > Key: CASSANDRA-12215 > URL: https://issues.apache.org/jira/browse/CASSANDRA-12215 > Project: Cassandra > Issue Type: Bug > Components: Compaction > Environment: Cassandra 3.0.8, cqlsh 5.0.1 > Reporter: Hau Phan > Fix For: 3.0.x > > > Running 3.0.8 on a single standalone node with cqlsh 5.0.1, the keyspace RF = > 1 and class SimpleStrategy. > Attempting to run a 'select * from <table>' and receiving this error: > ReadFailure: code=1300 [Replica(s) failed to execute read] message="Operation > failed - received 0 responses and 1 failures" info={'failures': 1, > 'received_responses': 0, 'required_responses': 1, 'consistency': 'ONE'} > Cassandra system.log prints this: > {code} > ERROR [CompactionExecutor:5] 2016-07-15 13:42:13,219 CassandraDaemon.java:201 > - Exception in thread Thread[CompactionExecutor:5,1,main] > java.lang.NullPointerException: null > at > org.apache.cassandra.db.transform.UnfilteredRows.isEmpty(UnfilteredRows.java:58) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.db.partitions.PurgeFunction.applyToPartition(PurgeFunction.java:64) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.db.partitions.PurgeFunction.applyToPartition(PurgeFunction.java:24) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.db.transform.BasePartitions.hasNext(BasePartitions.java:96) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.db.compaction.CompactionIterator.hasNext(CompactionIterator.java:226) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.db.compaction.CompactionTask.runMayThrow(CompactionTask.java:177) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.db.compaction.CompactionTask.executeInternal(CompactionTask.java:78) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.db.compaction.AbstractCompactionTask.execute(AbstractCompactionTask.java:60) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > org.apache.cassandra.db.compaction.CompactionManager$BackgroundCompactionCandidate.run(CompactionManager.java:263) > ~[apache-cassandra-3.0.8.jar:3.0.8] > at > java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) > ~[na:1.8.0_65] > at java.util.concurrent.FutureTask.run(FutureTask.java:266) > ~[na:1.8.0_65] > at > java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) > ~[na:1.8.0_65] > at > java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) > [na:1.8.0_65] > at java.lang.Thread.run(Thread.java:745) [na:1.8.0_65] > {code} > Doing a sstabledump -d shows a few rows with the column value of > "<tombstone>", telling me compaction doesn't seem to be working correctly. > # nodetool compactionstats > pending tasks: 1 > attempting to run a compaction gets: > # nodetool compact <table> <cf> > error: null > -- StackTrace -- > java.lang.NullPointerException > at > org.apache.cassandra.db.transform.UnfilteredRows.isEmpty(UnfilteredRows.java:58) > at > org.apache.cassandra.db.partitions.PurgeFunction.applyToPartition(PurgeFunction.java:64) > at > org.apache.cassandra.db.partitions.PurgeFunction.applyToPartition(PurgeFunction.java:24) > at > org.apache.cassandra.db.transform.BasePartitions.hasNext(BasePartitions.java:96) > at > org.apache.cassandra.db.compaction.CompactionIterator.hasNext(CompactionIterator.java:226) > at > org.apache.cassandra.db.compaction.CompactionTask.runMayThrow(CompactionTask.java:177) > at > org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28) > at > org.apache.cassandra.db.compaction.CompactionTask.executeInternal(CompactionTask.java:78) > at > org.apache.cassandra.db.compaction.AbstractCompactionTask.execute(AbstractCompactionTask.java:60) > at > org.apache.cassandra.db.compaction.CompactionManager$8.runMayThrow(CompactionManager.java:606) > at > org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28) > at > java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) > at java.util.concurrent.FutureTask.run(FutureTask.java:266) > at > java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) > at > java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) > at java.lang.Thread.run(Thread.java:745) > Since the table is pretty small, I can do a copy to, truncate the table, and > copy from, and the table is fine. But my concern is if compaction fails to > remove those rows, and the table will eventually be very large in a > production environment, the copy, truncate, and copy will no longer be an > option. -- This message was sent by Atlassian JIRA (v6.3.4#6332)