Hi German, I have applied patch 1810 and 1805 against trunk revision 1574686 (recent revision against which 1810 patch build succeeded). But observing following error in the zookeeper log on the new node joining quorum:
2014-03-10 21:11:25,126 [myid:1] - INFO [WorkerSender[myid=1]:QuorumCnxManager@195] - Have smaller server identifier, so dropping the connection: (3, 1) 2014-03-10 21:11:25,127 [myid:1] - INFO [/169.254.44.1:3888 :QuorumCnxManager$Listener@540] - Received connection request / 169.254.44.3:51507 2014-03-10 21:11:25,193 [myid:1] - ERROR [WorkerReceiver[myid=1]:NIOServerCnxnFactory$1@92] - Thread Thread[WorkerReceiver[myid=1],5,main] died java.lang.OutOfMemoryError: Java heap space at org.apache.zookeeper.server.quorum.FastLeaderElection$Messenger$WorkerReceiver.run(FastLeaderElection.java:273) at java.lang.Thread.run(Unknown Source) Followed by these messages getting printed repeatedly: 2014-03-10 21:11:25,328 [myid:1] - INFO [QuorumPeer[myid=1]/0:0:0:0:0:0:0:0:2181:FastLeaderElection@900] - Notification time out: 400 2014-03-10 21:11:25,729 [myid:1] - INFO [QuorumPeer[myid=1]/0:0:0:0:0:0:0:0:2181:FastLeaderElection@900] - Notification time out: 800 2014-03-10 21:11:26,530 [myid:1] - INFO [QuorumPeer[myid=1]/0:0:0:0:0:0:0:0:2181:FastLeaderElection@900] - Notification time out: 1600 2014-03-10 21:11:28,131 [myid:1] - INFO [QuorumPeer[myid=1]/0:0:0:0:0:0:0:0:2181:FastLeaderElection@900] - Notification time out: 3200 2014-03-10 21:11:31,332 [myid:1] - INFO [QuorumPeer[myid=1]/0:0:0:0:0:0:0:0:2181:FastLeaderElection@900] - Notification time out: 6400 Thanks & Reagrds, Deepak On Wed, Mar 5, 2014 at 11:50 AM, Deepak Jagtap <[email protected]>wrote: > Hi, > > I have applied only 1805 patch, not 1810. > And upgrade is from 3.5.0.1458648 to 3.5.0.1562289 (not from 3.4.5). > It was failing very consistently in our environment, and after 1805 patch > it went smoothly. > > Regards, > Deepak > > > > > > On Wed, Mar 5, 2014 at 7:36 AM, German Blanco < > [email protected]> wrote: > >> Hello, >> >> do you mean ZOOKEEPER-1810 patch? >> That one alone doesn't solve the problem. On the other hand, the problem >> doesn't happen always, so after a rolling start it might get solved. >> We need 1818 as well, but it is easier to go step by step and get 1810 in >> trunk first. >> I hope that as soon as 3.4.6 is out this might get some attention. >> >> Regards, >> >> German. >> >> >> On Wed, Mar 5, 2014 at 2:17 AM, Deepak Jagtap <[email protected] >> >wrote: >> >> > Hi, >> > >> > Please ignore the previous comment, I used wrong jar file and hence >> rolling >> > upgrade failed. >> > After applying patch for bug on zookeeper-3.5.0.1562289 >> > revision, rolling upgrade went fine. >> > >> > I have patched in house zookeeper version, but it would be convenient >> if we >> > apply patch on trunk and use the latest trunk. >> > Please advise if I can apply the patch on the trunk and test it for you. >> > >> > Thanks & Regards, >> > Deepak >> > >> > >> > On Tue, Mar 4, 2014 at 12:09 PM, Deepak Jagtap <[email protected] >> > >wrote: >> > >> > > Hi German, >> > > >> > > I tried applying patch for 1805 but problem still persists. >> > > Following are the notification messages logged repeatedly by the node >> > > which fails to join the quorum: >> > > >> > > >> > > 2014-03-04 20:00:54,398 [myid:2] - INFO >> > > [QuorumPeer[myid=2]/0:0:0:0:0:0:0:0:2181:FastLeaderElection@837] - >> > > Notification time out: 51200 >> > > 2014-03-04 20:00:54,400 [myid:2] - INFO >> > > [WorkerReceiver[myid=2]:FastLeaderElection@605] - Notification: 2 >> > > (n.leader), 0x0 (n.zxid), 0x1 (n.round), LOOKING (n.state), 2 (n.sid), >> > 0x0 >> > > (n.peerEPoch), LOOKING (my state)1 (n.config version) >> > > 2014-03-04 20:00:54,401 [myid:2] - INFO >> > > [WorkerReceiver[myid=2]:FastLeaderElection@605] - Notification: 3 >> > > (n.leader), 0x100003e84 (n.zxid), 0x2 (n.round), FOLLOWING (n.state), >> 1 >> > > (n.sid), 0x1 (n.peerEPoch), LOOKING (my state)1 (n.config version) >> > > 2014-03-04 20:00:54,403 [myid:2] - INFO >> > > [WorkerReceiver[myid=2]:FastLeaderElection@605] - Notification: 3 >> > > (n.leader), 0x100003e84 (n.zxid), 0xffffffffffffffff (n.round), >> LEADING >> > > (n.state), 3 (n.sid), 0x2 (n.peerEPoch), LOOKING (my state)1 (n.config >> > > version) >> > > >> > > >> > > >> > > Patch for 1732 is already included in the trunk. >> > > >> > > >> > > Thanks & Regards, >> > > Deepak >> > > >> > > >> > > On Fri, Feb 28, 2014 at 2:58 PM, Deepak Jagtap < >> [email protected] >> > >wrote: >> > > >> > >> Hi Flavio, German, >> > >> >> > >> Since this fix is critical for zookeeper rolling upgrade is it ok if >> I >> > >> apply this patch to 3.5.0 trunk? >> > >> Is it straightforward to apply this patch to trunk? >> > >> >> > >> Thanks & Regards, >> > >> Deepak >> > >> >> > >> >> > >> On Wed, Feb 26, 2014 at 11:46 AM, Deepak Jagtap < >> > [email protected]>wrote: >> > >> >> > >>> Thanks German! >> > >>> Just wondering is there any chance that this patch may be applied to >> > >>> trunk in near future? >> > >>> If it's fine with you guys, I would be more than happy to apply the >> > >>> fixes (from 3.4.5) to trunk and test them. >> > >>> >> > >>> Thanks & Regards, >> > >>> Deepak >> > >>> >> > >>> >> > >>> On Wed, Feb 26, 2014 at 1:29 AM, German Blanco < >> > >>> [email protected]> wrote: >> > >>> >> > >>>> Hello Deepak, >> > >>>> >> > >>>> due to ZOOKEEPER-1732 and then ZOOKEEPER-1805, there are some >> cases in >> > >>>> which an ensemble can be formed so that it doesn't allow any other >> > >>>> zookeeper server to join. >> > >>>> This has been fixed in branch 3.4, but it hasn't been fixed in >> trunk >> > >>>> yet. >> > >>>> Check if the Notifications sent around contain different values for >> > the >> > >>>> vote in the members of the ensemble. >> > >>>> If you force a new election (e.g. by killing the leader) I guess >> > >>>> everything >> > >>>> should work normally, but don't take my word for it. >> > >>>> Flavio should know more about this. >> > >>>> >> > >>>> Cheers, >> > >>>> >> > >>>> German. >> > >>>> >> > >>>> >> > >>>> On Wed, Feb 26, 2014 at 4:04 AM, Deepak Jagtap < >> > [email protected] >> > >>>> >wrote: >> > >>>> >> > >>>> > Hi, >> > >>>> > >> > >>>> > I replacing one of the zookeeper server from 3 node quorum. >> > >>>> > Initially all zookeeper serves were running 3.5.0.1515976 >> version. >> > >>>> > I successfully replaced Node3 with newer version 3.5.0.1551730. >> > >>>> > When I am trying to replace Node2 with the same zookeeper >> version. >> > >>>> > I couldn't start zookeeper server on Node2 as it is continuously >> > >>>> stuck in >> > >>>> > leader election loop printing following messages: >> > >>>> > >> > >>>> > 2014-02-26 02:45:23,709 [myid:3] - INFO >> > >>>> > [QuorumPeer[myid=3]/0:0:0:0:0:0:0:0:2181:FastLeaderElection@837] >> - >> > >>>> > Notification time out: 60000 >> > >>>> > 2014-02-26 02:45:23,710 [myid:3] - INFO >> > >>>> > [WorkerSender[myid=3]:QuorumCnxManager@195] - Have smaller >> server >> > >>>> > identifier, so dropping the connection: (5, 3) >> > >>>> > 2014-02-26 02:45:23,712 [myid:3] - INFO >> > >>>> > [WorkerReceiver[myid=3]:FastLeaderElection@605] - >> Notification: 3 >> > >>>> > (n.leader), 0x0 (n.zxid), 0x1 (n.round), LOOKING (n.state), 3 >> > >>>> (n.sid), 0x0 >> > >>>> > (n.peerEPoch), LOOKING (my state)1 (n.config version) >> > >>>> > >> > >>>> > >> > >>>> > Network connections and configuration of the node being upgraded >> are >> > >>>> fine. >> > >>>> > The other 2 nodes in the quorum are fine and serving the request. >> > >>>> > >> > >>>> > Any idea what might be causing this? >> > >>>> > >> > >>>> > Thanks & Regards, >> > >>>> > Deepak >> > >>>> > >> > >>>> >> > >>> >> > >>> >> > >> >> > > >> > >> > >
