[
https://issues.apache.org/jira/browse/YARN-5304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15817376#comment-15817376
]
Vrushali C commented on YARN-5304:
----------------------------------
[~jrottinghuis], [~sjlee0] and I had an offline discussion about this.
Summarizing the discussion here:
- Objective is to make things easier for the simple user who does not know
about HBase to get going with ATSv2 .
- Have a sample HBase config with default settings supplied with timeline
service that enables settings for timeline service v2. Ensure that
documentation has steps clarifying use of this HBase config for the HBase
cluster setup.
We discussed about having different HBase config files for the server side and
for clients. The server side HBase config can be made simple enough if it’s a
standalone deployment, say in case we are bringing up a HBase setup on the RM
node itself.
We discussed about providing a yarn-daemon command that can spin up a ATSv2
HBase backend using the sample HBase config supplied with the code. But this is
not feasible since this would mean ensuring the pids for HBase daemons are
handled by the yarn command. Also, HBase jars are needed to be made available,
which is outside of YARN/ATSv2 in any case.
In case of more complex deployment scenarios, like we have, say if there exists
an HBase cluster for ATSv2 data separate from other HBase clusters, then we
need a way to have different HBase configs such that there is a way to connect
from an application on a particular compute node on a hadoop cluster to two
different HBase clusters. One connection is for writing timeline service data
and another for the application to read/write from/to the other HBase cluster
for it’s own purpose, which [~jrottinghuis] addressed in YARN-5265.
> Ship single node HBase config option with single startup command
> ----------------------------------------------------------------
>
> Key: YARN-5304
> URL: https://issues.apache.org/jira/browse/YARN-5304
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: timelineserver
> Affects Versions: YARN-2928
> Reporter: Joep Rottinghuis
> Assignee: Vrushali C
> Labels: YARN-5355, yarn-5355-merge-blocker
>
> For small to medium Hadoop deployments we should make it dead-simple to use
> the timeline service v2. We should have a single command to launch and stop
> the timelineservice back-end for the default HBase implementation.
> A default config with all the values should be packaged that launches all the
> needed daemons (on the RM node) with a single command with all the
> recommended settings.
> Having a timeline admin command, perhaps an init command might be needed, or
> perhaps the timeline service can even auto-detect that and create tables,
> deploy needed coprocessors etc.
> The overall purpose is to ensure nobody needs to be an HBase expert to get
> this going. For those cluster operators with HBase experience, they can
> choose their own more sophisticated deployment.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
