zookeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Flavio Junqueira <fpjunque...@yahoo.com>
Subject Re: zookeeper quorum falling apart with continuous leader election
Date Wed, 12 Feb 2014 14:47:54 GMT
It sounds like LE is completing periodically, but the servers are not being able to complete
the synchronization step. We are also getting this connection refused exception when the follower
is trying to connect. This is what I spotted for the follower:

2014-02-10 18:54:04,414 [myid:234] - INFO  [QuorumPeer[myid=234]/0:0:0:0:0:0:0:0:2181:Follower@65]
2014-02-10 18:54:04,415 [myid:234] - WARN  [QuorumPeer[myid=234]/0:0:0:0:0:0:0:0:2181:Learner@239]
- Unexpected exception, tries=0, connecting to
java.net.ConnectException: Connection refused
	at java.net.PlainSocketImpl.socketConnect(Native Method)
	at java.net.AbstractPlainSocketImpl.doConnect(Unknown Source)
	at java.net.AbstractPlainSocketImpl.connectToAddress(Unknown Source)
	at java.net.AbstractPlainSocketImpl.connect(Unknown Source)
	at java.net.SocksSocketImpl.connect(Unknown Source)
	at java.net.Socket.connect(Unknown Source)
	at org.apache.zookeeper.server.quorum.Learner.connectToLeader(Learner.java:231)
	at org.apache.zookeeper.server.quorum.Follower.followLeader(Follower.java:73)
	at org.apache.zookeeper.server.quorum.QuorumPeer.run(QuorumPeer.java:936)

and this:

2014-02-10 18:55:05,508 [myid:234] - INFO  [QuorumPeer[myid=234]/0:0:0:0:0:0:0:0:2181:Learner@442]
- Learner received UPTODATE message
2014-02-10 18:55:05,508 [myid:234] - WARN  [QuorumPeer[myid=234]/0:0:0:0:0:0:0:0:2181:Follower@92]
- Exception when following the leader
java.net.SocketException: Broken pipe
	at java.net.SocketOutputStream.socketWrite0(Native Method)
	at java.net.SocketOutputStream.socketWrite(Unknown Source)
	at java.net.SocketOutputStream.write(Unknown Source)
	at java.io.BufferedOutputStream.flushBuffer(Unknown Source)
	at java.io.BufferedOutputStream.flush(Unknown Source)
	at org.apache.zookeeper.server.quorum.Learner.writePacket(Learner.java:145)
	at org.apache.zookeeper.server.quorum.Learner.syncWithLeader(Learner.java:477)
	at org.apache.zookeeper.server.quorum.Follower.followLeader(Follower.java:85)
	at org.apache.zookeeper.server.quorum.QuorumPeer.run(QuorumPeer.java:936)

On the leader side, we have this:

2014-02-10 19:48:03,705 [myid:235] - INFO  [LearnerHandler-/]
- Synchronizing with Follower sid: 234 maxCommittedLog=0x4afe00000001 minCommittedLog=0x4afe00000001
2014-02-10 19:48:03,705 [myid:235] - WARN  [LearnerHandler-/]
- Unhandled proposal scenario
2014-02-10 19:48:03,705 [myid:235] - INFO  [LearnerHandler-/]
- Sending SNAP
2014-02-10 19:48:03,705 [myid:235] - INFO  [LearnerHandler-/]
- Sending snapshot last zxid of peer is 0x4afd00000001  zxid of leader is 0x4aff00000000sent
zxid of db as 0x4afe00000001
2014-02-10 19:48:03,724 [myid:235] - WARN  [LearnerHandler-/]
- Commiting zxid 0x4aff00000000 from / not first!

There are a couple of odd warnings there. Just to confirm, the node missing in the logs is
the one with the bad disk, right?


On 12 Feb 2014, at 02:26, Deepak Jagtap <deepak.jagtap@maxta.com> wrote:

> Hi ,
> I have 3 node zookeeper quorum on my setup.
> We came across a situation where one of the zk server in the cluster went
> down
> due to bad disk.
> We observed that leader election keeps running in loop (starts, completes
> and again starts). The loop repeats every couple of minutes.
> Even restarting zookeeper server on both nodes doesn't help recovering from
> this loop.
> Network connection looks fine though, as I could telnet leader election
> port and ssh from one node to other.
> zookeeper client on each node is using "" as quorum string
> for connecting to server, therefore if local zookeeper server is down
> client app is dead.
> I have uploaded zookeeper.log for both nodes at following link:
> https://dl.dropboxusercontent.com/u/36429721/zkSupportLog.tar.gz
> Any idea what might be wrong with the quorum? Please note that restarting
> zookeeper server on both nodes doesn't help to recover from this situations.
> Thanks & Regards,
> Deepak

View raw message