[HBASE-24794] hbase.rowlock.wait.duration should not be <= 0

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.4.0, 1.5.0, 2.2.0, 2.3.0, 1.6.0
• Fix Version/s: 3.0.0-alpha-1, 2.3.1, 1.7.0, 2.4.0, 1.4.14, 2.2.6
• Component/s: regionserver
• Assignee: Sean Busbey
• Reporter: Sean Busbey
• Created: 29/Jul/20 16:50
• Updated: 01/Aug/20 00:32
• Resolved: 30/Jul/20 20:22

Description

had a cluster fail after upgrade from hbase 1 because all writes to meta failed.

master started in maintenance mode looks like (RS hosting meta in non-maint would look similar starting with HRegion.doBatchMutate):

2020-07-28 17:52:56,553 WARN org.apache.hadoop.hbase.regionserver.HRegion: Failed getting lock, row=some_user_table
java.io.IOException: Timed out waiting for lock for row: some_user_table in region 1588230740
        at org.apache.hadoop.hbase.regionserver.HRegion.getRowLockInternal(HRegion.java:5863)
        at org.apache.hadoop.hbase.regionserver.HRegion$BatchOperation.lockRowsAndBuildMiniBatch(HRegion.java:3322)
        at org.apache.hadoop.hbase.regionserver.HRegion.doMiniBatchMutate(HRegion.java:4018)
        at org.apache.hadoop.hbase.regionserver.HRegion.batchMutate(HRegion.java:3992)
        at org.apache.hadoop.hbase.regionserver.HRegion.batchMutate(HRegion.java:3923)
        at org.apache.hadoop.hbase.regionserver.HRegion.batchMutate(HRegion.java:3914)
        at org.apache.hadoop.hbase.regionserver.HRegion.batchMutate(HRegion.java:3928)
        at org.apache.hadoop.hbase.regionserver.HRegion.doBatchMutate(HRegion.java:4255)
        at org.apache.hadoop.hbase.regionserver.HRegion.put(HRegion.java:3047)
        at org.apache.hadoop.hbase.regionserver.RSRpcServices.mutate(RSRpcServices.java:2827)
        at org.apache.hadoop.hbase.client.ClientServiceCallable.doMutate(ClientServiceCallable.java:55)
        at org.apache.hadoop.hbase.client.HTable$3.rpcCall(HTable.java:538)
        at org.apache.hadoop.hbase.client.HTable$3.rpcCall(HTable.java:533)
        at org.apache.hadoop.hbase.client.RegionServerCallable.call(RegionServerCallable.java:127)
        at org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.callWithRetries(RpcRetryingCallerImpl.java:107)
        at org.apache.hadoop.hbase.client.HTable.put(HTable.java:542)
        at org.apache.hadoop.hbase.MetaTableAccessor.put(MetaTableAccessor.java:1339)
        at org.apache.hadoop.hbase.MetaTableAccessor.putToMetaTable(MetaTableAccessor.java:1329)
        at org.apache.hadoop.hbase.MetaTableAccessor.updateTableState(MetaTableAccessor.java:1672)
        at org.apache.hadoop.hbase.MetaTableAccessor.updateTableState(MetaTableAccessor.java:1112)
        at org.apache.hadoop.hbase.master.TableStateManager.fixTableStates(TableStateManager.java:296)
        at org.apache.hadoop.hbase.master.TableStateManager.start(TableStateManager.java:269)
        at org.apache.hadoop.hbase.master.HMaster.finishActiveMasterInitialization(HMaster.java:1004)
        at org.apache.hadoop.hbase.master.HMaster.startActiveMasterManager(HMaster.java:2274)
        at org.apache.hadoop.hbase.master.HMaster.lambda$run$0(HMaster.java:583)
        at java.lang.Thread.run(Thread.java:745)

logging roughly 6k times /second.

failure was caused by a change in behavior for hbase.rowlock.wait.duration in HBASE-17210 (so 1.4.0+, 2.0.0+). Prior to that change setting the config <= 0 meant that row locks would succeed only if they were immediately available. After the change we fail the lock attempt without checking the lock at all.

workaround: set hbase.rowlock.wait.duration to a small positive number, e.g. 1, if you want row locks to fail quickly.

Issue Links

• is broken by: HBASE-17210 Set timeout on trying rowlock according to client's RPC timeout
• links to: GitHub Pull Request #2174