[
https://issues.apache.org/jira/browse/AMBARI-19720?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15838854#comment-15838854
]
Sumit Mohanty commented on AMBARI-19720:
----------------------------------------
Reference impl from run.sh
{code}
cat /var/lib/ambari-agent/tmp/llap-slider2017-01-24_21-26-05/run.sh
#!/bin/bash -e
BASEDIR=$(dirname $0)
slider stop llap0
slider destroy llap0 --force || slider destroy llap0
slider install-package --name LLAP --package $BASEDIR/llap-24Jan2017.zip
--replacepkg
slider create llap0 --resources $BASEDIR/resources.json --template
$BASEDIR/appConfig.json --queue llap
{code}
> when restarting HSI, Ambari should not run slider stop and slider destroy
> (LLAP script already does that)
> ---------------------------------------------------------------------------------------------------------
>
> Key: AMBARI-19720
> URL: https://issues.apache.org/jira/browse/AMBARI-19720
> Project: Ambari
> Issue Type: Bug
> Components: stacks
> Affects Versions: 2.5.0
> Reporter: Sumit Mohanty
> Assignee: Sumit Mohanty
> Priority: Critical
> Fix For: 2.5.0
>
>
> These commands are pure overhead, because spinning up slider and connecting
> to things is slow. On some test cluster it took ~12sec to do 100ms of noop
> work.
> LLAP script already runs them and will eventually avoid the overhead by
> calling Slider APIs from an existing JVM. However we just noticed when
> restarting HSI (but not when restarting LLAP only), Ambari also runs these
> commands.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
