[
https://issues.apache.org/jira/browse/AMBARI-13930?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15011851#comment-15011851
]
Jayush Luniya commented on AMBARI-13930:
----------------------------------------
Branch-2.1:
commit d01b1d80671ae6bfa4db2ed8d9ab23b04ba31f38
Author: Jayush Luniya <[email protected]>
Date: Wed Nov 18 11:48:50 2015 -0800
AMBARI-13930: Express Upgrade: Failure at Save Cluster State because
cluster_version=OUT_OF_SYNC (jluniya)
Conflicts:
ambari-server/src/main/java/org/apache/ambari/server/api/handlers/CreateHandler.java
> Express Upgrade: Failure at Save Cluster State because
> cluster_version=OUT_OF_SYNC
> -----------------------------------------------------------------------------------
>
> Key: AMBARI-13930
> URL: https://issues.apache.org/jira/browse/AMBARI-13930
> Project: Ambari
> Issue Type: Bug
> Components: ambari-server
> Affects Versions: 2.1.3
> Reporter: Jayush Luniya
> Assignee: Jayush Luniya
> Priority: Critical
> Fix For: 2.1.3
>
>
> ambari-server --hash
> cc77797be2d645d1f1c798617eb837ab04204fdc
> # Deploy HDP 2.2.6.0 (Services: HDFS, YARN, ZOOKEEPER)
> # Datanode on jay-213-4, jay-213-5.
> # Express Upgrade HDP 2.2.6.0 -> HDP 2.3.0.0
> # Intentionally cause datanode on jay-213-5 to fail. Ignore and Proceed on
> datanode failure.
> # Finalize upgrade precheck fails reporting datanode on jay-213-5.
> # Pause Upgrade, restart datanode successfully
> # Resume Upgrade
> # Finalize Upgrade fails at "Save Cluster State"
> {noformat}
> The cluster stack version state OUT_OF_SYNC is not allowed to transition
> directly into CURRENT
> {noformat}
> Ambari Server Log
> {noformat}
> 13 Nov 2015 07:07:49,152 WARN [qtp-ambari-agent-459] ClusterImpl:1262 - have
> a mix of CURRENT and INSTALLING|INSTALLED|UPGRADING|UPGRADED host versions,
> returning OUT_OF_SYNC as cluster version. Host version states:
> {UPGRADED=[jay-213-3.c.pramod-thangali.internal],
> INSTALLED=[jay-213-5.c.pramod-thangali.internal,
> jay-213-4.c.pramod-thangali.internal, jay-213-2.c.pramod-thangali.internal,
> jay-213-1.c.pramod-thangali.internal]}
> {noformat}
> This is possible if a node has a single component (Zookeeper instance) and is
> the 1st instance to be upgraded. Before any of the other hosts is updated
> from INSTALLED to some other state, the host state for this host is already
> been changed to UPGRADED leading to cluster version going OUT_OF_SYNC.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
