[jira] [Created] (CASSANDRA-15898) cassandra 3.11.4 deadlock

["john doe (Jira)" <ji...@apache.org>]

john doe created CASSANDRA-15898:
------------------------------------

             Summary: cassandra 3.11.4 deadlock
                 Key: CASSANDRA-15898
                 URL: https://issues.apache.org/jira/browse/CASSANDRA-15898
             Project: Cassandra
          Issue Type: Bug
            Reporter: john doe


We are running apache-cassandra-3.11.4, 10 node cluster with -Xms32G -Xmx32G -Xmn8G using CMS.
after running couple of days one of the node become unresponsive and threaddump (jstack -F) shows deadlock.


Found one Java-level deadlock:
=============================

"Native-Transport-Requests-144": waiting to lock Monitor@0x00007cd5142e4d08 (Object@0x00007f6e00348268, a java/io/ExpiringCache),
 which is held by "CompactionExecutor:115134"
"CompactionExecutor:115134": waiting to lock Monitor@0x00007f6bcaf130f8 (Object@0x00007f6dff31faa0, a ch/qos/logback/core/joran/spi/ConfigurationWatchList),
 which is held by "Native-Transport-Requests-144"

Found a total of 1 deadlock.

I have seen this couple of time now with different nodes with following in system.log

IndexSummaryRedistribution.java:77 - Redistributing index summaries
 NoSpamLogger.java:91 - Maximum memory usage reached (536870912), cannot allocate chunk of 1048576

also lookin in gc log there has not been a ParNew collection for last 10hrs, only CMS collections.

1739842.375: [GC (CMS Final Remark) [YG occupancy: 2712269 K (7549760 K)]
1739842.375: [Rescan (parallel) , 0.0614157 secs]
1739842.437: [weak refs processing, 0.0000994 secs]
1739842.437: [class unloading, 0.0231076 secs]
1739842.460: [scrub symbol table, 0.0061049 secs]
1739842.466: [scrub string table, 0.0043847 secs][1 CMS-remark: 17696837K(25165824K)] 20409107K(32715584K), 0.0953750 secs] [Times: user=2.95 sys=0.00, real=0.09 secs]
1739842.471: [CMS-concurrent-sweep-start]
1739848.572: [CMS-concurrent-sweep: 6.101/6.101 secs] [Times: user=6.13 sys=0.00, real=6.10 secs]
1739848.573: [CMS-concurrent-reset-start]
1739848.645: [CMS-concurrent-reset: 0.072/0.072 secs] [Times: user=0.08 sys=0.00, real=0.08 secs]
1739858.653: [GC (CMS Initial Mark) [1 CMS-initial-mark: 17696837K(25165824K)] 
20409111K(32715584K), 0.0584838 secs] [Times: user=2.68 sys=0.00, real=0.06 secs]
1739858.713: [CMS-concurrent-mark-start]
1739860.496: [CMS-concurrent-mark: 1.784/1.784 secs] [Times: user=84.77 sys=0.00, real=1.79 secs]
1739860.497: [CMS-concurrent-preclean-start]
1739860.566: [CMS-concurrent-preclean: 0.070/0.070 secs] [Times: user=0.07 sys=0.00, real=0.07 secs]
1739860.567: [CMS-concurrent-abortable-preclean-start]CMS: abort preclean due to time
1739866.333: [CMS-concurrent-abortable-preclean: 5.766/5.766 secs] [Times: user=5.80 sys=0.00, real=5.76 secs]

Java HotSpot(TM) 64-Bit Server VM (25.162-b12) for linux-amd64 JRE (1.8.0_162-b12)
Memory: 4k page, physical 792290076k(2780032k free), swap 16777212k(16693756k free)

CommandLine flags:
-XX:+AlwaysPreTouch
-XX:CICompilerCount=15
-XX:+CMSClassUnloadingEnabled
-XX:+CMSEdenChunksRecordAlways
-XX:CMSInitiatingOccupancyFraction=40
-XX:+CMSParallelInitialMarkEnabled
-XX:+CMSParallelRemarkEnabled
-XX:CMSWaitDuration=10000
-XX:ConcGCThreads=50
-XX:+CrashOnOutOfMemoryError
-XX:GCLogFileSize=10485760
-XX:+HeapDumpOnOutOfMemoryError
-XX:InitialHeapSize=34359738368
-XX:InitialTenuringThreshold=1
-XX:+ManagementServer
-XX:MaxHeapSize=34359738368
-XX:MaxNewSize=8589934592
-XX:MaxTenuringThreshold=1
-XX:MinHeapDeltaBytes=196608
-XX:NewSize=8589934592
-XX:NumberOfGCLogFiles=10
-XX:OldPLABSize=16
-XX:OldSize=25769803776
-XX:OnOutOfMemoryError=kill -9 %p
-XX:ParallelGCThreads=50
-XX:+PerfDisableSharedMem
-XX:+PrintGC
-XX:+PrintGCDetails
-XX:+PrintGCTimeStamps
-XX:+ResizeTLAB
-XX:StringTableSize=1000003
-XX:SurvivorRatio=8
-XX:ThreadPriorityPolicy=42
-XX:ThreadStackSize=256
-XX:-UseBiasedLocking
-XX:+UseCMSInitiatingOccupancyOnly
-XX:+UseConcMarkSweepGC
-XX:+UseCondCardMark
-XX:+UseFastUnorderedTimeStamps
-XX:+UseGCLogFileRotation
-XX:+UseNUMA
-XX:+UseNUMAInterleaving
-XX:+UseParNewGC
-XX:+UseTLAB
-XX:+UseThreadPriorities



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org