【原创】大叔问题定位分享（1）HBase RegionServer频繁挂掉

最近hbase集群很多region server挂掉，查看其中一个RegionServer1日志发现，17:17:14挂的时候服务器压力很大，有大量的responseTooSlow，也有不少gc，但是当时内存还有很多剩余，不是因为oom被kill

2018-03-13T17:17:13.372+0800: [GC (Allocation Failure) 2018-03-13T17:17:13.372+0800: [ParNew: 3280066K->256481K(3762880K), 0.0429718 secs] 23536952K->20513367K(32801920K), 0.0432932 secs] [Times: user=1.92 sys=0.01, real=0.04 secs]

Heap

par new generation   total 3762880K, used 1416770K [0x00007f2f4c000000, 0x00007f305f990000, 0x00007f305f990000)

eden space 3010368K,  38% used [0x00007f2f4c000000, 0x00007f2f92d182c0, 0x00007f3003bd0000)

from space 752512K,  34% used [0x00007f3003bd0000, 0x00007f30136487c0, 0x00007f3031ab0000)

to   space 752512K,   0% used [0x00007f3031ab0000, 0x00007f3031ab0000, 0x00007f305f990000)

concurrent mark-sweep generation total 29039040K, used 19905973K [0x00007f305f990000, 0x00007f374c000000, 0x00007f374c000000)

Metaspace       used 49530K, capacity 50072K, committed 55492K, reserved 57344K

region server挂掉的原因是tryRegionServerReport时发现被标记为dead server，抛出YouAreDeadException，然后HRegionServer.run方法会退出，意味着region server进程退出

2018-03-13 17:17:08,159 FATAL [regionserver/RegionServer1/RegionServer1:16020] regionserver.HRegionServer: ABORTING region server RegionServer1,16020,1519805863844: org.apache.hadoop.hbase.YouAreDeadException: Server REPORT rejected; currently processing RegionServer1,16020,1519805863844 as dead server

org.apache.hadoop.hbase.YouAreDeadException: org.apache.hadoop.hbase.YouAreDeadException: Server REPORT rejected; currently processing RegionServer1,16020,1519805863844 as dead server

at org.apache.hadoop.hbase.master.ServerManager.checkIsDead(ServerManager.java:434)

at org.apache.hadoop.hbase.master.ServerManager.regionServerReport(ServerManager.java:339)

at org.apache.hadoop.hbase.master.MasterRpcServices.regionServerReport(MasterRpcServices.java:339)

at org.apache.hadoop.hbase.protobuf.generated.RegionServerStatusProtos$RegionServerStatusService$2.callBlockingMethod(Reg

ionServerStatusProtos.java:8617)

at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:2196)

at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:112)

at org.apache.hadoop.hbase.ipc.RpcExecutor.consumerLoop(RpcExecutor.java:133)

at org.apache.hadoop.hbase.ipc.RpcExecutor$1.run(RpcExecutor.java:108)

at java.lang.Thread.run(Thread.java:745)

at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)

at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)

at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)

at java.lang.reflect.Constructor.newInstance(Constructor.java:422)

at org.apache.hadoop.ipc.RemoteException.instantiateException(RemoteException.java:106)

at org.apache.hadoop.ipc.RemoteException.unwrapRemoteException(RemoteException.java:95)

at org.apache.hadoop.hbase.protobuf.ProtobufUtil.getRemoteException(ProtobufUtil.java:330)

at org.apache.hadoop.hbase.regionserver.HRegionServer.tryRegionServerReport(HRegionServer.java:1153)

at org.apache.hadoop.hbase.regionserver.HRegionServer.run(HRegionServer.java:962)

被标记为dead server是因为zk session expired

2018-03-13 17:17:08,128 INFO  [regionserver/RegionServer1/RegionServer1:16020-SendThread(ZKServer1:2181)] zookeeper.ClientCnxn: Client session timed out, have not heard from server in 41499ms for sessionid 0x161d1433ba86c84, closing socket connection and attempting reconnect

2018-03-13 17:17:08,445 FATAL [main-EventThread] regionserver.HRegionServer: ABORTING region server RegionServer1,16020,1519805863844: regionserver:16020-0x35fe18af217e685, quorum=ZKServer1:2181, baseZNode=/hbase regionserver:16020-0x35fe18af217e685 received expired from ZooKeeper, aborting

在17:17:08超时41499ms是因为17:16:38的时候有一个full gc耗时29s

2018-03-13T17:16:38.268+0800: [GC (Allocation Failure)

2018-03-13T17:16:38.268+0800: [ParNew (promotion failed): 3762880K->3762880K(3762880K), 3.2398750 secs]

2018-03-13T17:16:41.508+0800: [CMS2018-03-13T17:16:47.078+0800: [CMS-concurrent-sweep: 10.829/17.135 secs] [Times: user=161.97 sys=2.65, real=17.13 secs]

(concurrent mode failure): 28187725K->21036514K(29039040K), 26.5884371 secs] 31874835K->21036514K(32801920K), [Metaspace: 49335K->49335K(57344K)], 29.8297421 secs] [Times: user=40.45 sys=0.14, real=29.82 secs]

但是hbase-site.xml中配置session超时是180s，为什么41s就超时

<property>

<name>zookeeper.session.timeout</name>

<value>180000</value>

</property>

建立zk session时的negotiated timeout = 40000，也就是40s，为什么配置的是180s，实际是40s？

2018-03-13 17:15:11,522 INFO  [main-SendThread(ZKServer1:2181)] zookeeper.ClientCnxn: Session establishment complet

e on server ZKServer1/ZKServer1:2181, sessionid = 0x35fe18af217e685, negotiated timeout = 40000

2018-03-13 17:15:11,526 INFO  [regionserver/RegionServer1/RegionServer1:16020-SendThread(ZKServer1:2181)]

zookeeper.ClientCnxn: Session establishment complete on server RegionServer1/RegionServer1:2181, sessionid = 0x161d1433b

a86c84, negotiated timeout = 40000

zk中建立连接后输出

            LOG.info("Session establishment complete on server "

                    + clientCnxnSocket.getRemoteSocketAddress()

                    + ", sessionid = 0x" + Long.toHexString(sessionId)

                    + ", negotiated timeout = " + negotiatedSessionTimeout

                    + (isRO ? " (READ-ONLY mode)" : ""));

negotiatedSessionTimeout是在onConnected中被回调

        void onConnected(int _negotiatedSessionTimeout, long _sessionId,

                byte[] _sessionPasswd, boolean isRO) throws IOException {

回调是在ClientCnxnSocket中，取得是ConnectResponse.getTimeout()

        ConnectResponse conRsp = new ConnectResponse();

        conRsp.deserialize(bbia, "connect");

                  sendThread.onConnected(conRsp.getTimeOut(), this.sessionId,

                conRsp.getPasswd(), isRO);

这里取的是ServerCnxn中的sessionTimeout，这个值在ZooKeeperServer中做初始化

        int minSessionTimeout = getMinSessionTimeout();

        if (sessionTimeout < minSessionTimeout) {

            sessionTimeout = minSessionTimeout;

        }

        int maxSessionTimeout = getMaxSessionTimeout();

        if (sessionTimeout > maxSessionTimeout) {

            sessionTimeout = maxSessionTimeout;

        }

可以看到会根据minSessionTimeout和maxSessionTimeout限制

    public int getMaxSessionTimeout() {

        return maxSessionTimeout == -1 ? tickTime * 20 : maxSessionTimeout;

    }

由于zookeeper服务器tickTime设置的是2000ms，所以maxSessionTimeout默认会被设置为40000ms，所以解决这个问题需要修改zk的maxSessionTimeout；