[ https://issues.apache.org/jira/browse/ATLAS-2786?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16672331#comment-16672331 ]
ASF subversion and git services commented on ATLAS-2786: -------------------------------------------------------- Commit 5b5788e9c5ae1c043c4bf52cedbf8b586aa93d90 in atlas's branch refs/heads/branch-1.0 from [~mad...@apache.org] [ https://git-wip-us.apache.org/repos/asf?p=atlas.git;h=5b5788e ] ATLAS-2786: updated HBase client to use all zookeeper entries (cherry picked from commit 2f39428313136b765dbab42a1ee2e6521695d690) > [Atlas HA] Zookeeper connection issue > ------------------------------------- > > Key: ATLAS-2786 > URL: https://issues.apache.org/jira/browse/ATLAS-2786 > Project: Atlas > Issue Type: Bug > Affects Versions: 0.8.1, 0.8.2, 1.0.0 > Reporter: Apoorv Naik > Assignee: Apoorv Naik > Priority: Critical > Fix For: 0.8.3, 1.1.0, 2.0.0 > > Attachments: > 0001-ATLAS-2786-Honor-ZK-server-ensemble-provided-in-conf.patch, > ATLAS-2786-2.patch > > > The curator client right now only connects to the very first zookeeper server > regardless of the number of zookeeper servers provided in the configuration. > > Due to this behavior if the ZK (that Atlas connected to during startup) goes > down then Atlas fails to serve further request while constantly trying to > reconnect to the dead ZK server. > > This patch makes the change in configuration parsing to allow fallback to > different ZK servers in case the primary goes down. -- This message was sent by Atlassian JIRA (v7.6.3#76005)