[HBASE-13317] Region server reportForDuty stuck looping if there is a master change

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.0.0, 0.98.12, 2.0.0
• Fix Version/s: 1.0.1, 1.1.0, 0.98.12, 2.0.0
• Component/s: regionserver
• Hadoop Flags: Reviewed
• Assignee: Jerry He
• Reporter: Jerry He
• Created: 23/Mar/15 16:42
• Updated: 28/Apr/15 19:19
• Resolved: 01/Apr/15 06:06

Description

During cluster startup, region server reportForDuty gets stuck looping if there is a master change.

2015-03-22 11:15:16,186 INFO  [regionserver60020] regionserver.HRegionServer: reportForDuty to master=bigaperf274,60000,1427045883965 with port=60020, startcode=1427048115174
2015-03-22 11:15:16,272 WARN  [regionserver60020] regionserver.HRegionServer: error telling master we are up
com.google.protobuf.ServiceException: java.net.ConnectException: Connection refused
	at org.apache.hadoop.hbase.ipc.RpcClient.callBlockingMethod(RpcClient.java:1678)
	at org.apache.hadoop.hbase.ipc.RpcClient$BlockingRpcChannelImplementation.callBlockingMethod(RpcClient.java:1719)
	at org.apache.hadoop.hbase.protobuf.generated.RegionServerStatusProtos$RegionServerStatusService$BlockingStub.regionServerStartup(RegionServerStatusProtos.java:8277)
	at org.apache.hadoop.hbase.regionserver.HRegionServer.reportForDuty(HRegionServer.java:2137)
	at org.apache.hadoop.hbase.regionserver.HRegionServer.run(HRegionServer.java:896)
	at java.lang.Thread.run(Thread.java:745)
2015-03-22 11:15:16,274 WARN  [regionserver60020] regionserver.HRegionServer: reportForDuty failed; sleeping and then retrying.
2015-03-22 11:15:19,274 INFO  [regionserver60020] regionserver.HRegionServer: reportForDuty to master=bigaperf273,60000,1427048108439 with port=60020, startcode=1427048115174
2015-03-22 11:15:19,275 WARN  [regionserver60020] regionserver.HRegionServer: error telling master we are up
com.google.protobuf.ServiceException: java.net.ConnectException: Connection refused
	at org.apache.hadoop.hbase.ipc.RpcClient.callBlockingMethod(RpcClient.java:1678)
	at org.apache.hadoop.hbase.ipc.RpcClient$BlockingRpcChannelImplementation.callBlockingMethod(RpcClient.java:1719)
	at org.apache.hadoop.hbase.protobuf.generated.RegionServerStatusProtos$RegionServerStatusService$BlockingStub.regionServerStartup(RegionServerStatusProtos.java:8277)
	at org.apache.hadoop.hbase.regionserver.HRegionServer.reportForDuty(HRegionServer.java:2137)
	at org.apache.hadoop.hbase.regionserver.HRegionServer.run(HRegionServer.java:896)
	at java.lang.Thread.run(Thread.java:745)
2015-03-22 11:15:19,276 WARN  [regionserver60020] regionserver.HRegionServer: reportForDuty failed; sleeping and then retrying.
2015-03-22 11:15:22,276 INFO  [regionserver60020] regionserver.HRegionServer: reportForDuty to master=bigaperf273,60000,1427048108439 with port=60020, startcode=1427048115174
2015-03-22 11:15:22,296 DEBUG [regionserver60020] regionserver.HRegionServer: Master is not running yet
2015-03-22 11:15:22,296 WARN  [regionserver60020] regionserver.HRegionServer: reportForDuty failed; sleeping and then retrying.
2015-03-22 11:15:25,296 INFO  [regionserver60020] regionserver.HRegionServer: reportForDuty to master=bigaperf273,60000,1427048108439 with port=60020, startcode=1427048115174
2015-03-22 11:15:25,299 DEBUG [regionserver60020] regionserver.HRegionServer: Master is not running yet
2015-03-22 11:15:25,299 WARN  [regionserver60020] regionserver.HRegionServer: reportForDuty failed; sleeping and then retrying.
2015-03-22 11:15:28,299 INFO  [regionserver60020] regionserver.HRegionServer: reportForDuty to master=bigaperf273,60000,1427048108439 with port=60020, startcode=1427048115174
2015-03-22 11:15:28,302 DEBUG [regionserver60020] regionserver.HRegionServer: Master is not running yet
2015-03-22 11:15:28,302 WARN  [regionserver60020] regionserver.HRegionServer: reportForDuty failed; sleeping and then retrying.

What happended is the region server first got master=bigaperf274,60000,1427045883965. Before it was able to report successfully, the maser changed to bigaperf273,60000,1427048108439.
We were supposed to open a new connection to the new master. But we never did, looping and trying to old address forever.

Issue Links

• duplicates: HBASE-12813 Reporting region in transition shouldn't loop forever
• relates to: HBASE-13345 Fix LocalHBaseCluster so that different region server impl can be used for different slaves