[HBASE-14266] RegionServers have a lock contention of Configuration.getProps

• Type: Improvement
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 1.2.0, 1.3.0, 2.0.0
• Component/s: regionserver

Environment

hbase-0.98.6-cdh5.3.1

• Hadoop Flags: Reviewed
• Assignee: Andrew Kyle Purtell
• Reporter: Toshihiro Suzuki
• Created: 20/Aug/15 07:24
• Updated: 17/Jun/22 18:55
• Resolved: 24/Oct/15 22:39

Description

Here's an extract from thread dump of the RegionServer of my cluster:

...
Thread 267 (RW.default.readRpcServer.handler=184,queue=15,port=60020):
  State: BLOCKED
  Blocked count: 204028
  Waited count: 9702639
  Blocked on org.apache.hadoop.conf.Configuration@5a5e3da
  Blocked by 250 (RW.default.readRpcServer.handler=167,queue=18,port=60020)
  Stack:
    org.apache.hadoop.conf.Configuration.getProps(Configuration.java:2250)
    org.apache.hadoop.conf.Configuration.get(Configuration.java:861)
    org.apache.hadoop.conf.Configuration.getTrimmed(Configuration.java:880)
    org.apache.hadoop.conf.Configuration.getBoolean(Configuration.java:1281)
    org.apache.hadoop.hbase.regionserver.StoreScanner.<init>(StoreScanner.java:138)
    org.apache.hadoop.hbase.regionserver.StoreScanner.<init>(StoreScanner.java:157)
    org.apache.hadoop.hbase.regionserver.HStore.createScanner(HStore.java:1804)
    org.apache.hadoop.hbase.regionserver.HStore.getScanner(HStore.java:1794)
    org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.<init>(HRegion.java:3852)
    org.apache.hadoop.hbase.regionserver.HRegion.instantiateRegionScanner(HRegion.java:1952)
    org.apache.hadoop.hbase.regionserver.HRegion.getScanner(HRegion.java:1938)
    org.apache.hadoop.hbase.regionserver.HRegion.getScanner(HRegion.java:1915)
    org.apache.hadoop.hbase.regionserver.HRegion.get(HRegion.java:4872)
    org.apache.hadoop.hbase.regionserver.HRegion.get(HRegion.java:4847)
    org.apache.hadoop.hbase.regionserver.HRegionServer.get(HRegionServer.java:2918)
    org.apache.hadoop.hbase.protobuf.generated.ClientProtos$ClientService$2.callBlockingMethod(ClientProtos.java:29921)
    org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2031)
    org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:108)
    org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:116)
    org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:96)
...

There are such many threads in the thread dump.
I think that RegionServers have a lock contention which causes performance issue.

Sub-Tasks

• Cache Configuration used during scanner creation: Closed