使用happybase訪問HBase出現Broken pipe問題---兩個“驚天”大bug

來源
使用happybase通過thrift接口向HBase讀取、寫入數據時，出現Broken pipe的錯誤。排查步驟：

1、查看hbase的日志：

Java HotSpot(TM) 64-Bit Server VM warning: Using incremental CMS is deprecated and will likely be removed in a future release
17/05/12 18:08:41 INFO util.VersionInfo: HBase 1.2.0-cdh6.10.1
17/05/12 18:08:41 INFO util.VersionInfo: Source code repository file:///data/jenkins/workspace/generic-package-centos64-7-0/topdir/BUILD/hbase-1.2.0-cdh6.10.1 revision=Unknown
17/05/12 18:08:41 INFO util.VersionInfo: Compiled by jenkins on Mon Mar 20 02:46:09 PDT 2017
17/05/12 18:08:41 INFO util.VersionInfo: From source with checksum c6d9864e1358df7e7f39d39a40338b4e
17/05/12 18:08:41 INFO thrift.ThriftServerRunner: Using default thrift server type
17/05/12 18:08:41 INFO thrift.ThriftServerRunner: Using thrift server type threadpool
17/05/12 18:08:42 WARN impl.MetricsConfig: Cannot locate configuration: tried hadoop-metrics2-hbase.properties,hadoop-metrics2.properties
17/05/12 18:08:42 INFO impl.MetricsSystemImpl: Scheduled snapshot period at 10 second(s).
17/05/12 18:08:42 INFO impl.MetricsSystemImpl: HBase metrics system started
17/05/12 18:08:42 INFO mortbay.log: Logging to org.slf4j.impl.Log4jLoggerAdapter(org.mortbay.log) via org.mortbay.log.Slf4jLog
17/05/12 18:08:42 INFO http.HttpRequestLog: Http request log for http.requests.thrift is not defined
17/05/12 18:08:42 INFO http.HttpServer: Added global filter 'safety' (class=org.apache.hadoop.hbase.http.HttpServer$QuotingInputFilter)
17/05/12 18:08:42 INFO http.HttpServer: Added global filter 'clickjackingprevention' (class=org.apache.hadoop.hbase.http.ClickjackingPreventionFilter)
17/05/12 18:08:42 INFO http.HttpServer: Added filter static_user_filter (class=org.apache.hadoop.hbase.http.lib.StaticUserWebFilter$StaticUserFilter) to context thrift
17/05/12 18:08:42 INFO http.HttpServer: Added filter static_user_filter (class=org.apache.hadoop.hbase.http.lib.StaticUserWebFilter$StaticUserFilter) to context static
17/05/12 18:08:42 INFO http.HttpServer: Added filter static_user_filter (class=org.apache.hadoop.hbase.http.lib.StaticUserWebFilter$StaticUserFilter) to context logs
17/05/12 18:08:42 INFO http.HttpServer: Jetty bound to port 9095
17/05/12 18:08:42 INFO mortbay.log: jetty-6.1.26.cloudera.4
17/05/12 18:08:42 WARN mortbay.log: Can't reuse /tmp/Jetty_0_0_0_0_9095_thrift____.vqpz9l, using /tmp/Jetty_0_0_0_0_9095_thrift____.vqpz9l_5120175032480185058
17/05/12 18:08:43 INFO mortbay.log: Started SelectChannelConnector@0.0.0.0:9095
17/05/12 18:08:43 INFO thrift.ThriftServerRunner: starting TBoundedThreadPoolServer on /0.0.0.0:9090 with readTimeout 300000ms; min worker threads=128, max worker threads=1000, max queued requests=1000
.
.
.
/05/08 15:05:51 INFO zookeeper.RecoverableZooKeeper: Process identifier=hconnection-0x645132bf connecting to ZooKeeper ensemble=cdh-master-slave1:2181,cdh-slave2:2181,cdh-slave3:2181
17/05/08 15:05:51 INFO zookeeper.ZooKeeper: Initiating client connection, connectString=cdh-master-slave1:2181,cdh-slave2:2181,cdh-slave3:2181 sessionTimeout=60000 watcher=hconnection-0x64513-master-slave1:2181,cdh-slave2:2181,cdh-slave3:2181, baseZNode=/hbase
17/05/08 15:05:51 INFO zookeeper.ClientCnxn: Opening socket connection to server cdh-slave3/192.168.10.219:2181. Will not attempt to authenticate using SASL (unknown error)
17/05/08 15:05:51 INFO zookeeper.ClientCnxn: Socket connection established, initiating session, client: /192.168.10.23:43170, server: cdh-slave3/192.168.10.219:2181
17/05/08 15:05:51 INFO zookeeper.ClientCnxn: Session establishment complete on server cdh-slave3/192.168.10.219:2181, sessionid = 0x35bd74a77802148, negotiated timeout = 60000
[caitinggui@cdh-master-slave1 example]$ 17/05/08 15:32:50 INFO client.ConnectionManager$HConnectionImplementation: Closing zookeeper sessionid=0x35bd74a77802148
17/05/08 15:32:51 INFO zookeeper.ZooKeeper: Session: 0x35bd74a77802148 closed
17/05/08 15:32:51 INFO zookeeper.ClientCnxn: EventThread shut down
17/05/08 15:38:53 INFO zookeeper.RecoverableZooKeeper: Process identifier=hconnection-0xb876351 connecting to ZooKeeper ensemble=cdh-master-slave1:2181,cdh-slave2:2181,cdh-slave3:2181
17/05/08 15:38:53 INFO zookeeper.ZooKeeper: Initiating client connection, connectString=cdh-master-slave1:2181,cdh-slave2:2181,cdh-slave3:2181 sessionTimeout=60000 watcher=hconnection-0xb8763510x0, quorum=cdh-master-slave1:2181,cdh-slave2:2181,cdh-slave3:2181, baseZNode=/hbase
17/05/08 15:38:53 INFO zookeeper.ClientCnxn: Opening socket connection to server cdh-master-slave1/192.168.10.23:2181. Will not attempt to authenticate using SASL (unknown error)
17/05/08 15:38:53 INFO zookeeper.ClientCnxn: Socket connection established, initiating session, client: /192.168.10.23:35526, server: cdh-master-slave1/192.168.10.23:2181
17/05/08 15:38:53 INFO zookeeper.ClientCnxn: Session establishment complete on server cdh-master-slave1/192.168.10.23:2181, sessionid = 0x15ba3ddc6cc90d4, negotiated timeout = 60000

初步推斷是hbase設置了某個超時時間，導致連接斷開

2、查看官方文檔，但是沒有發現很有意義的timeout參數

3、Google相似問題

查看相似的內容：

Uploaded image for project: 'HBase' 
HBaseHBASE-14926
Hung ThriftServer; no timeout on read from client; if client crashes, worker thread gets stuck reading
Agile Board
 Export
Details
Type: Bug
Status:RESOLVED
Priority: Major
Resolution: Fixed
Affects Version/s:
2.0.0, 1.2.0, 1.1.2, 1.3.0, 1.0.3, 0.98.16
Fix Version/s:
2.0.0, 1.2.0, 1.3.0, 0.98.17
Component/s:
Thrift
Labels:None
Hadoop Flags:
Reviewed
Release Note: Adds a timeout to server read from clients. Adds new configs hbase.thrift.server.socket.read.timeout for setting read timeout on server socket in milliseconds. Default is 60000;
Description
Thrift server is hung. All worker threads are doing this:
"thrift-worker-0" daemon prio=10 tid=0x00007f0bb95c2800 nid=0xf6a7 runnable [0x00007f0b956e0000]
   java.lang.Thread.State: RUNNABLE
        at java.net.SocketInputStream.socketRead0(Native Method)
        at java.net.SocketInputStream.read(SocketInputStream.java:152)
        at java.net.SocketInputStream.read(SocketInputStream.java:122)
        at java.io.BufferedInputStream.fill(BufferedInputStream.java:235)
        at java.io.BufferedInputStream.read1(BufferedInputStream.java:275)
        at java.io.BufferedInputStream.read(BufferedInputStream.java:334)
        - locked <0x000000066d859490> (a java.io.BufferedInputStream)
        at org.apache.thrift.transport.TIOStreamTransport.read(TIOStreamTransport.java:127)
        at org.apache.thrift.transport.TTransport.readAll(TTransport.java:84)
        at org.apache.thrift.transport.TFramedTransport.readFrame(TFramedTransport.java:129)
        at org.apache.thrift.transport.TFramedTransport.read(TFramedTransport.java:101)
        at org.apache.thrift.transport.TTransport.readAll(TTransport.java:84)
        at org.apache.thrift.protocol.TCompactProtocol.readByte(TCompactProtocol.java:601)
        at org.apache.thrift.protocol.TCompactProtocol.readMessageBegin(TCompactProtocol.java:470)
        at org.apache.thrift.TBaseProcessor.process(TBaseProcessor.java:27)
        at org.apache.hadoop.hbase.thrift.TBoundedThreadPoolServer$ClientConnnection.run(TBoundedThreadPoolServer.java:289)
        at org.apache.hadoop.hbase.thrift.CallQueue$Call.run(CallQueue.java:64)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
        at java.lang.Thread.run(Thread.java:745)
They never recover.
I don't have client side logs.
We've been here before: HBASE-4967 "connected client thrift sockets should have a server side read timeout" but this patch only got applied to fb branch (and thrift has changed since then)
ps：來源https://issues.apache.org/jira/browse/HBASE-14926

4、Google “hbase.thrift.server.socket.read.timeout”

可以看到一個網頁內容：

問題背景

測試環境是三臺服務器搭建的Hadoop分布式環境。Hadoop版本是：hadoop-2.7.3；Hbase-1.2.4； 
zookeeper-3.4.9。 
使用thrift c++接口向hbase中寫入數據，每次都是剛開始寫入正常，過一段時間就開始報錯。 
但之前使用的hbase-0.94.27版本就沒遇到過該問題，配置也相同，一直用的好好地。

thrift接口報錯

解決辦法

通過抓包可以看出，hbase server響應了RST包，導致連接中斷。 
通過 bin/hbase thrift start -threadpool命令可以readTimeout的設置為60s。

thriftpool

經過驗證卻是和這個設置有關，配置中沒有配置過該項，通過查看代碼發現60s是默認值，如果沒有配置即按照以該值為準。

因此在conf/hbase-site.xml中添加上配置即可：

<property>
         <name>hbase.thrift.server.socket.read.timeout</name>
         <value>6000000</value>
         <description>eg:milisecond</description>
</property>

ps：來源http://blog.csdn.net/wwlhz/article/details/56012053

所以添加參數后，重啟hbase thrift，發現問題解決

5、查看源碼，可以看到

#https://github.com/apache/hbase/blob/master/hbase-thrift/src/main/java/org/apache/hadoop/hbase/thrift/ThriftServerRunner.java

...
  public static final String THRIFT_SERVER_SOCKET_READ_TIMEOUT_KEY =
    "hbase.thrift.server.socket.read.timeout";
  public static final int THRIFT_SERVER_SOCKET_READ_TIMEOUT_DEFAULT = 60000;
.
.
.
      int readTimeout = conf.getInt(THRIFT_SERVER_SOCKET_READ_TIMEOUT_KEY,
          THRIFT_SERVER_SOCKET_READ_TIMEOUT_DEFAULT);
      TServerTransport serverTransport = new TServerSocket(
          new TServerSocket.ServerSocketTransportArgs().
              bindAddr(new InetSocketAddress(listenAddress, listenPort)).
              backlog(backlog).
              clientTimeout(readTimeout));

問題解決~~~

6、然而問題解決了嗎？

實際上還是有問題，一段時間發現連續scan大概20多分鐘后，連接又被斷開了，又是一次艱難的搜索，發現是hbase該版本自帶的問題，它將所有連接（不管有沒有在使用）都默認為idle的狀態，然后有個hbase.thrift.connection.max-idletime的配置，所以我將此項配置為31104000（一年），如果是在CDH中，應該在管理頁面配置，如圖：

遇到問題一般步驟：
技術進步型：
1、查看日志，查看報錯的地方，初步定位問題
2、查看官方文檔
3、Google相似的問題，或者查看源碼去定位問題

快速解決問題型：
1、查看日志，查看報錯的地方，初步定位問題
2、Google相似問題
3、查看官方文檔，或者查看源碼

參考：

• [1]HBase thrift/thrift2 使用指南