[ https://issues.apache.org/jira/browse/ZOOKEEPER-2867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16121780#comment-16121780 ]
Jun Rao commented on ZOOKEEPER-2867: ------------------------------------ [~hanm], in this case, session 25cd1e82c110001 was created by a Kafka broker. The broker registers an ephemeral node in ZK on startup. That Kafka broker went down around Aug 2, 23:16:11. The expectation is that its ZK session will be closed and the registered ephemeral node will be removed. Another Kafka broker (the controller) has a child watcher on the ephemeral node's parent path. It will be notified about the broker change and process it accordingly. The problem is that the controller never received such notification. This led the Kafka cluster to a bad state. You were saying that the following entry doesn't necessarily mean that the ephemeral node is actually removed since it could be restored. Is there any logging indicating the restore? {code:java} August 2nd 2017, 23:41:29.149 DEBUG org.apache.zookeeper.server.DataTree Deleting ephemeral node /cp15/brokers/ids/0 for session 0x25cd1e82c110001 {code} > an expired ZK session can be re-established > ------------------------------------------- > > Key: ZOOKEEPER-2867 > URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2867 > Project: ZooKeeper > Issue Type: Bug > Affects Versions: 3.4.10 > Reporter: Jun Rao > Attachments: zk.0.formatted, zk.1.formatted > > > Not sure if this is a real bug, but I found an instance when a ZK client > seems to be able to renew a session already expired by the ZK server. > From ZK server log, session 25cd1e82c110001 was expired at 22:04:39. > {code:java} > June 27th 2017, 22:04:39.000 INFO > org.apache.zookeeper.server.ZooKeeperServer Expiring session > 0x25cd1e82c110001, timeout of 12000ms exceeded > June 27th 2017, 22:04:39.001 DEBUG > org.apache.zookeeper.server.quorum.Leader Proposing:: > sessionid:0x25cd1e82c110001 type:closeSession cxid:0x0 zxid:0x200000fc4 > txntype:-11 reqpath:n/a > June 27th 2017, 22:04:39.001 INFO > org.apache.zookeeper.server.PrepRequestProcessor Processed session > termination for sessionid: 0x25cd1e82c110001 > June 27th 2017, 22:04:39.001 DEBUG > org.apache.zookeeper.server.quorum.CommitProcessor Processing request:: > sessionid:0x25cd1e82c110001 type:closeSession cxid:0x0 zxid:0x200000fc4 > txntype:-11 reqpath:n/a > June 27th 2017, 22:05:20.324 INFO > org.apache.zookeeper.server.quorum.Learner Revalidating client: > 0x25cd1e82c110001 > June 27th 2017, 22:05:20.324 INFO > org.apache.zookeeper.server.ZooKeeperServer Client attempting to renew > session 0x25cd1e82c110001 at /100.96.5.6:47618 > June 27th 2017, 22:05:20.325 INFO > org.apache.zookeeper.server.ZooKeeperServer Established session > 0x25cd1e82c110001 with negotiated timeout 12000 for client /100.96.5.6:47618 > {code} > From ZK client's log, it was able to renew the expired session on 22:05:20. > {code:java} > June 27th 2017, 22:05:18.590 INFO org.apache.zookeeper.ClientCnxn Client > session timed out, have not heard from server in 4485ms for sessionid > 0x25cd1e82c110001, closing socket connection and attempting reconnect 0 > June 27th 2017, 22:05:18.590 WARN org.apache.zookeeper.ClientCnxn Client > session timed out, have not heard from server in 4485ms for sessionid > 0x25cd1e82c110001 0 > June 27th 2017, 22:05:19.325 WARN org.apache.zookeeper.ClientCnxn SASL > configuration failed: javax.security.auth.login.LoginException: No JAAS > configuration section named 'Client' was found in specified JAAS > configuration file: '/opt/confluent/etc/kafka/server_jaas.conf'. Will > continue connection to Zookeeper server without SASL authentication, if > Zookeeper server allows it. 0 > June 27th 2017, 22:05:19.326 INFO org.apache.zookeeper.ClientCnxn Opening > socket connection to server 100.65.188.168/100.65.188.168:2181 0 > June 27th 2017, 22:05:20.324 INFO org.apache.zookeeper.ClientCnxn Socket > connection established to 100.65.188.168/100.65.188.168:2181, initiating > session 0 > June 27th 2017, 22:05:20.327 INFO org.apache.zookeeper.ClientCnxn Session > establishment complete on server 100.65.188.168/100.65.188.168:2181, > sessionid = 0x25cd1e82c110001, negotiated timeout = 12000 0 > {code} -- This message was sent by Atlassian JIRA (v6.4.14#64029)