[HBASE-10272] Cluster becomes nonoperational if the node hosting the active Master AND ROOT/META table goes offline

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 0.96.1, 0.94.15
• Fix Version/s: 0.98.0, 0.94.16, 0.96.2, 0.99.0
• Component/s: IPC/RPC
• Hadoop Flags: Reviewed
• Assignee: Aditya Kishore
• Reporter: Aditya Kishore
• Created: 03/Jan/14 01:03
• Updated: 21/Jan/14 01:31
• Resolved: 06/Jan/14 20:15

Description

Since HBASE-6364, HBase client caches a connection failure to a server and any subsequent attempt to connect to the server throws a FailedServerException

Now if a node which hosted the active Master AND ROOT/META table goes offline, the newly anointed Master's initial attempt to connect to the dead region server will fail with NoRouteToHostException which it handles but since on second attempt crashes with FailedServerException

Here is the log from one such occurance

2013-11-20 10:58:00,161 FATAL org.apache.hadoop.hbase.master.HMaster: Master server abort: loaded coprocessors are: []
2013-11-20 10:58:00,161 FATAL org.apache.hadoop.hbase.master.HMaster: Unhandled exception. Starting shutdown.
org.apache.hadoop.hbase.ipc.HBaseClient$FailedServerException: This server is in the failed servers list: xxx02/192.168.1.102:60020
        at org.apache.hadoop.hbase.ipc.HBaseClient$Connection.setupIOstreams(HBaseClient.java:425)
        at org.apache.hadoop.hbase.ipc.HBaseClient.getConnection(HBaseClient.java:1124)
        at org.apache.hadoop.hbase.ipc.HBaseClient.call(HBaseClient.java:974)
        at org.apache.hadoop.hbase.ipc.WritableRpcEngine$Invoker.invoke(WritableRpcEngine.java:86)
        at $Proxy9.getProtocolVersion(Unknown Source)
        at org.apache.hadoop.hbase.ipc.WritableRpcEngine.getProxy(WritableRpcEngine.java:138)
        at org.apache.hadoop.hbase.ipc.HBaseRPC.waitForProxy(HBaseRPC.java:208)
        at org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation.getHRegionConnection(HConnectionManager.java:1335)
        at org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation.getHRegionConnection(HConnectionManager.java:1294)
        at org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation.getHRegionConnection(HConnectionManager.java:1281)
        at org.apache.hadoop.hbase.catalog.CatalogTracker.getCachedConnection(CatalogTracker.java:506)
        at org.apache.hadoop.hbase.catalog.CatalogTracker.getMetaServerConnection(CatalogTracker.java:383)
        at org.apache.hadoop.hbase.catalog.CatalogTracker.waitForMeta(CatalogTracker.java:445)
        at org.apache.hadoop.hbase.catalog.CatalogTracker.waitForMetaServerConnection(CatalogTracker.java:464)
        at org.apache.hadoop.hbase.catalog.CatalogTracker.verifyMetaRegionLocation(CatalogTracker.java:624)
        at org.apache.hadoop.hbase.master.HMaster.assignRootAndMeta(HMaster.java:684)
        at org.apache.hadoop.hbase.master.HMaster.finishInitialization(HMaster.java:560)
        at org.apache.hadoop.hbase.master.HMaster.run(HMaster.java:376)
        at java.lang.Thread.run(Thread.java:662)
2013-11-20 10:58:00,162 INFO org.apache.hadoop.hbase.master.HMaster: Aborting
2013-11-20 10:58:00,162 INFO org.apache.hadoop.ipc.HBaseServer: Stopping server on 60000

Each of the backup master will crash with same error and restarting them will have the same effect. Once this happens, the cluster will remain in-operational until the node with region server is brought online (or the Zookeeper node containing the root region server and/or META entry from the ROOT table is deleted).