[
https://issues.apache.org/jira/browse/HDFS-14059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16687093#comment-16687093
]
Plamen Jeliazkov commented on HDFS-14059:
-----------------------------------------
New finding. Let's call it (4):
With `dfs.ha.automatic-failover.enabled=true` still set, I am noticing that
when I manually transition a Standby->Observer (that has ZKFC co-located), the
ZKFC will automatically try to convert the Observer back to Standby mode. Logs
end up looking like this:
{code}
2018-11-14 12:29:00,466 ERROR org.apache.hadoop.ha.ZKFailoverController: Local
service NameNode at
instance-3.pp-devcos-myhadoop.us-central1.gcp.dev.paypalinc.com/10.176.1.207:8030
has changed the serviceState to observer. Expected was standby. Quitting
election marking fencing necessary.
2018-11-14 12:29:00,466 INFO org.apache.hadoop.ha.ActiveStandbyElector:
Yielding from election
2018-11-14 12:29:00,468 INFO org.apache.zookeeper.ZooKeeper: Session:
0x1000acb2b350012 closed
2018-11-14 12:29:00,468 INFO org.apache.zookeeper.ClientCnxn: EventThread shut
down for session: 0x1000acb2b350012
2018-11-14 12:29:01,469 INFO org.apache.zookeeper.ZooKeeper: Initiating client
connection,
connectString=instance-3.pp-devcos-myhadoop.us-central1.gcp.dev.paypalinc.com:2181
sessionTimeout=10000
watcher=org.apache.hadoop.ha.ActiveStandbyElector$WatcherWithClientRef@2992f4e4
2018-11-14 12:29:01,471 INFO org.apache.zookeeper.ClientCnxn: Opening socket
connection to server
instance-3.pp-devcos-myhadoop.us-central1.gcp.dev.paypalinc.com/10.176.1.207:2181.
Will not attempt to authenticate using SASL (unknown error)
2018-11-14 12:29:01,471 INFO org.apache.zookeeper.ClientCnxn: Socket connection
established to
instance-3.pp-devcos-myhadoop.us-central1.gcp.dev.paypalinc.com/10.176.1.207:2181,
initiating session
2018-11-14 12:29:01,474 INFO org.apache.zookeeper.ClientCnxn: Session
establishment complete on server
instance-3.pp-devcos-myhadoop.us-central1.gcp.dev.paypalinc.com/10.176.1.207:2181,
sessionid = 0x1000acb2b350013, negotiated timeout = 10000
2018-11-14 12:29:01,475 INFO org.apache.hadoop.ha.ActiveStandbyElector: Session
connected.
2018-11-14 12:29:01,479 INFO org.apache.hadoop.ha.ZKFailoverController: ZK
Election indicated that NameNode at
instance-3.pp-devcos-myhadoop.us-central1.gcp.dev.paypalinc.com/10.176.1.207:8030
should become standby
2018-11-14 12:29:01,503 INFO org.apache.hadoop.ha.ZKFailoverController:
Successfully transitioned NameNode at
instance-3.pp-devcos-myhadoop.us-central1.gcp.dev.paypalinc.com/10.176.1.207:8030
to standby state
{code}
With the ZKFC on the Standby killed, I am able to transition it to Observer and
able to create directories, files, and then list status, cat, etc., as usual.
It seems we need to make a decision on whether we want to support automatic
failover and go into ZKFC and, possibly, ZK states, or not support automatic
failover but still support ConfiguredFailoverProxyProvider.
> Test reads from standby on a secure cluster with Configured failover
> --------------------------------------------------------------------
>
> Key: HDFS-14059
> URL: https://issues.apache.org/jira/browse/HDFS-14059
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: test
> Reporter: Konstantin Shvachko
> Assignee: Plamen Jeliazkov
> Priority: Major
>
> Run standard HDFS tests to verify reading from ObserverNode on a secure HA
> cluster with {{ConfiguredFailoverProxyProvider}}.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
