[CASSANDRA-8964] SSTable count rises during compactions and max open files exceeded

• Type: Bug
• Status: Resolved
• Priority: Urgent
• Resolution: Cannot Reproduce
• Labels: lcs

Environment

Apache Cassandra 2.1.2
Centos 6
AWS EC2
i2.2xlarge

• Severity: Critical
• Since Version: 2.1.2
• Assignee: Unassigned
• Reporter: Anthony Fisk
• Created: 12/Mar/15 22:17
• Updated: 16/Apr/19 09:31
• Resolved: 24/Mar/15 14:30

Description

LCS compaction was not able to keep up with the prolonged insert load on one of our tables called log, resulting in 2,185 SSTables for that table and 1,779 pending compactions all together during a test we were running.

We stopped our load, unthrottled compaction throughput, increased the concurrent compactors from 2 to 8, and let it compact the SSTables.

All was going well until the number of SSTables count for our log table got down to around 97, then began rising again until it had reached 758 SSTables 1.5 hours later... (we've been recording the cfstats output every half hour, attached)

Eventually we exceeded the number of open files:

ERROR [MemtableFlushWriter:286] 2015-03-12 13:44:36,748 CassandraDaemon.java:153 - Exception in thread Thread[MemtableFlushWriter:286,5,main]
java.lang.RuntimeException: java.io.FileNotFoundException: /mnt/cassandra/data/system/compactions_in_progress-55080ab05d9c388690a4acb25fe1f77b/system-compactions_in_progress-tmp-ka-6618-Index.db (Too many open files)
        at org.apache.cassandra.io.util.SequentialWriter.<init>(SequentialWriter.java:75) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.util.SequentialWriter.open(SequentialWriter.java:104) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.util.SequentialWriter.open(SequentialWriter.java:99) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.sstable.SSTableWriter$IndexWriter.<init>(SSTableWriter.java:552) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.sstable.SSTableWriter.<init>(SSTableWriter.java:134) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.db.Memtable$FlushRunnable.createFlushWriter(Memtable.java:390) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.db.Memtable$FlushRunnable.writeSortedContents(Memtable.java:329) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.db.Memtable$FlushRunnable.runWith(Memtable.java:313) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.util.DiskAwareRunnable.runMayThrow(DiskAwareRunnable.java:48) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at com.google.common.util.concurrent.MoreExecutors$SameThreadExecutorService.execute(MoreExecutors.java:297) ~[guava-16.0.jar:na]
        at org.apache.cassandra.db.ColumnFamilyStore$Flush.run(ColumnFamilyStore.java:1037) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) ~[na:1.7.0_51]
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) ~[na:1.7.0_51]
        at java.lang.Thread.run(Thread.java:744) ~[na:1.7.0_51]
Caused by: java.io.FileNotFoundException: /mnt/cassandra/data/system/compactions_in_progress-55080ab05d9c388690a4acb25fe1f77b/system-compactions_in_progress-tmp-ka-6618-Index.db (Too many open files)
        at java.io.RandomAccessFile.open(Native Method) ~[na:1.7.0_51]
        at java.io.RandomAccessFile.<init>(RandomAccessFile.java:241) ~[na:1.7.0_51]
        at org.apache.cassandra.io.util.SequentialWriter.<init>(SequentialWriter.java:71) ~[apache-cassandra-2.1.2.jar:2.1.2]
        ... 14 common frames omitted
ERROR [MemtableFlushWriter:286] 2015-03-12 13:44:36,750 JVMStabilityInspector.java:94 - JVM state determined to be unstable.  Exiting forcefully due to:
java.io.FileNotFoundException: /mnt/cassandra/data/system/compactions_in_progress-55080ab05d9c388690a4acb25fe1f77b/system-compactions_in_progress-tmp-ka-6618-Index.db (Too many open files)
        at java.io.RandomAccessFile.open(Native Method) ~[na:1.7.0_51]
        at java.io.RandomAccessFile.<init>(RandomAccessFile.java:241) ~[na:1.7.0_51]
        at org.apache.cassandra.io.util.SequentialWriter.<init>(SequentialWriter.java:71) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.util.SequentialWriter.open(SequentialWriter.java:104) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.util.SequentialWriter.open(SequentialWriter.java:99) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.sstable.SSTableWriter$IndexWriter.<init>(SSTableWriter.java:552) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.sstable.SSTableWriter.<init>(SSTableWriter.java:134) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.db.Memtable$FlushRunnable.createFlushWriter(Memtable.java:390) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.db.Memtable$FlushRunnable.writeSortedContents(Memtable.java:329) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.db.Memtable$FlushRunnable.runWith(Memtable.java:313) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.io.util.DiskAwareRunnable.runMayThrow(DiskAwareRunnable.java:48) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at com.google.common.util.concurrent.MoreExecutors$SameThreadExecutorService.execute(MoreExecutors.java:297) ~[guava-16.0.jar:na]
        at org.apache.cassandra.db.ColumnFamilyStore$Flush.run(ColumnFamilyStore.java:1037) ~[apache-cassandra-2.1.2.jar:2.1.2]
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) ~[na:1.7.0_51]
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) ~[na:1.7.0_51]
        at java.lang.Thread.run(Thread.java:744) ~[na:1.7.0_51]

I restarted Cassandra and the number of SSTables reduced to 81 before increasing again... Looking at lsof output, it seemed that the file handles for the log table were also continually increasing. It seemed that roughly half of them were for "tmp" files and the other half were for the directory of the table. Also of note, it seemed we only had one compactor running on the log table, rather than the 8 that we had seen before after changing the configuration and restarting.

Cassandra eventually crashed again with the same error when it reached 6,442 open files (seen in lsof). Without Cassandra running our system has 1,126 open files.

Issue Links

• is duplicated by: CASSANDRA-8939 Stack overflow when reading data ingested through SSTableLoader