[
https://issues.apache.org/jira/browse/TRAFODION-2663?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16067062#comment-16067062
]
Steve Varnau commented on TRAFODION-2663:
-----------------------------------------
The install is failing on HDP distro.
Due to error messages in HBase logs, I thought this was due to a timing change
in how quickly HBase comes up (master initilizing) and so I added a re-try in
the hbcheck script.
With more debugging, I found that it is actually a change in how quickly the
regionserver looks for it's bulkload staging directory in HDFS. Since
python-installer changes the configured path of the directory, but does not
create the HDFS directory until after it restarts HBase, the regionserver does
not find it and exits.
My changes had nothing to do with bulkload directory, but apparently reducing
the regionserver coprocessor classes changed the timing significantly enough to
get to the check before the HDFS commands ran.
> Simplify installation setting of HBase config parameters
> --------------------------------------------------------
>
> Key: TRAFODION-2663
> URL: https://issues.apache.org/jira/browse/TRAFODION-2663
> Project: Apache Trafodion
> Issue Type: Improvement
> Components: installer
> Reporter: Steve Varnau
> Assignee: Steve Varnau
>
> Affects both py-installer and ambari-installer.
> Due to Selva's work in [TRAFODION-2306], some parameters no longer need to be
> set in HBase config. The settings like hbase.coprocessor.region.classes &
> hbase.client.scanner.timeout.period that are now set in trafodion-site.xml
> file do not need to be set in HBase service-wide config by installers.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
