[HBASE-19163] "Maximum lock count exceeded" from region server's batch processing

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0-alpha-1, 2.0.0-alpha-3, 1.2.7
• Fix Version/s: 1.4.1, 1.3.3, 2.0.0
• Component/s: regionserver

Release Note

When there are many mutations against the same row in a batch, as each mutation will acquire a shared row lock, it will exceed the maximum shared lock count the java ReadWritelock supports (64k). Along with other optimization, the batch is divided into multiple possible minibatches. A new config is added to limit the maximum number of mutations in the minibatch.

   <property>
    <name>hbase.regionserver.minibatch.size</name>
    <value>20000</value>
   </property>
The default value is 20000.

• Assignee: Hua Xiang
• Reporter: Hua Xiang
• Created: 02/Nov/17 21:28
• Updated: 01/Feb/19 19:56
• Resolved: 19/Jan/18 19:29

Description

In one of use cases, we found the following exception and replication is stuck.

2017-10-25 19:41:17,199 WARN  [hconnection-0x28db294f-shared--pool4-t936] client.AsyncProcess: #3, table=foo, attempt=5/5 failed=262836ops, last exception: java.io.IOException: java.io.IOException: Maximum lock count exceeded
        at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2215)
        at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:109)
        at org.apache.hadoop.hbase.ipc.RpcExecutor$Handler.run(RpcExecutor.java:185)
        at org.apache.hadoop.hbase.ipc.RpcExecutor$Handler.run(RpcExecutor.java:165)
Caused by: java.lang.Error: Maximum lock count exceeded
        at java.util.concurrent.locks.ReentrantReadWriteLock$Sync.fullTryAcquireShared(ReentrantReadWriteLock.java:528)
        at java.util.concurrent.locks.ReentrantReadWriteLock$Sync.tryAcquireShared(ReentrantReadWriteLock.java:488)
        at java.util.concurrent.locks.AbstractQueuedSynchronizer.tryAcquireSharedNanos(AbstractQueuedSynchronizer.java:1327)
        at java.util.concurrent.locks.ReentrantReadWriteLock$ReadLock.tryLock(ReentrantReadWriteLock.java:871)
        at org.apache.hadoop.hbase.regionserver.HRegion.getRowLock(HRegion.java:5163)
        at org.apache.hadoop.hbase.regionserver.HRegion.doMiniBatchMutation(HRegion.java:3018)
        at org.apache.hadoop.hbase.regionserver.HRegion.batchMutate(HRegion.java:2877)
        at org.apache.hadoop.hbase.regionserver.HRegion.batchMutate(HRegion.java:2819)
        at org.apache.hadoop.hbase.regionserver.RSRpcServices.doBatchOp(RSRpcServices.java:753)
        at org.apache.hadoop.hbase.regionserver.RSRpcServices.doNonAtomicRegionMutation(RSRpcServices.java:715)
        at org.apache.hadoop.hbase.regionserver.RSRpcServices.multi(RSRpcServices.java:2148)
        at org.apache.hadoop.hbase.protobuf.generated.ClientProtos$ClientService$2.callBlockingMethod(ClientProtos.java:33656)
        at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2170)
        ... 3 more

While we are still examining the data pattern, it is sure that there are too many mutations in the batch against the same row, this exceeds the maximum 64k shared lock count and it throws an error and failed the whole batch.

There are two approaches to solve this issue.

1). Let's say there are mutations against the same row in the batch, we just need to acquire the lock once for the same row vs to acquire the lock for each mutation.
2). We catch the error and start to process whatever it gets and loop back.

With HBASE-17924, approach 1 seems easy to implement now.
Create the jira and will post update/patch when investigation moving forward.

Issue Links

• links to: Review Board (master)