[
https://issues.apache.org/jira/browse/SOLR-12163?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16420934#comment-16420934
]
Varun Thacker commented on SOLR-12163:
--------------------------------------
1. "Any shards or Solr instances" -> "All remaining Solr nodes in the cluster
that rely..."
2. To me this section "Running an Ensemble on a Single Server" looks like it
could be skipped ? If someone wants to run 3 ZK instances on one sever , we
firstly shouldn't be helping them and secondly they might as well run one
instance? And it just adds more docs and makes it complicated to read ?
3. One really important thing users miss out while configuring zoo.cfg is
setting autopurge.snapRetainCount and autopurge.purgeInterval . If you don't
see the latter then basically the snapshots never get deleted and users run of
out disk space. So let's add the following two params to zoo.cfg as well
autopurge.snapRetainCount=3
autopurge.purgeInterval=1
4. We need to create a file called zookeeper-env.sh under zookeeper_homne/conf
4.1 To configure ZooKeeper logging add the following to zookeeper-env.sh.
Please additionally configure log file size in conf/log4j.properties ( line 41 )
ZOO_LOG_DIR="/Users/varunthacker/zookeeper-3.4.9/logging"
ZOO_LOG4J_PROP="INFO,ROLLINGFILE"
4.2. To configure heap size and enable GC logging add the following to
zookeeper-env.sh:
SERVER_JVMFLAGS="-Xms2048m -Xmx2048m -verbose:gc -XX:+PrintHeapAtGC
-XX:+PrintGCDetails -XX:+PrintGCDateStamps -XX:+PrintGCTimeStamps
-XX:+PrintTenuringDistribution -XX:+PrintGCApplicationStoppedTime
-Xloggc:$ZOO_LOG_DIR/zookeeper_gc.log -XX:+UseGCLogFileRotation
-XX:NumberOfGCLogFiles=9 -XX:GCLogFileSize=20M"
FYI when i was testing this to see if it all works, I didn't see anything in
the zookeeper_gc.log.0.current file. I had to force a GC through jconsole and
then entries started to show up so don't be surprised if you don't see any gc
log entries on startup.
> Ref Guide: Improve Setting Up an External ZK Ensemble page
> ----------------------------------------------------------
>
> Key: SOLR-12163
> URL: https://issues.apache.org/jira/browse/SOLR-12163
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Components: documentation
> Reporter: Cassandra Targett
> Assignee: Cassandra Targett
> Priority: Major
> Fix For: 7.4
>
> Attachments: setting-up-an-external-zookeeper-ensemble.adoc
>
>
> I had to set up a ZK ensemble the other day for the first time in a while,
> and thought I'd test our docs on the subject while I was at it. I headed over
> to
> https://lucene.apache.org/solr/guide/setting-up-an-external-zookeeper-ensemble.html,
> and...Well, I still haven't gotten back to what I was trying to do, but I
> rewrote the entire page.
> The problem to me is that the page today is mostly a stripped down copy of
> the ZK Getting Started docs: walking through setting up a single ZK instance
> before introducing the idea of an ensemble and going back through the same
> configs again to update them for the ensemble.
> IOW, despite the page being titled "setting up an ensemble", it's mostly
> about not setting up an ensemble. That's at the end of the page, which itself
> focuses a bit heavily on the use case of running an ensemble on a single
> server (so, if you're counting...that's 3 use cases we don't want people to
> use discussed in detail on a page that's supposedly about _not_ doing any of
> those things).
> So, I took all of it and restructured the whole thing to focus primarily on
> the use case we want people to use: running 3 ZK nodes on different machines.
> Running 3 on one machine is still there, but noted in passing with the
> appropriate caveats. I've also added information about choosing to use a
> chroot, which AFAICT was only covered in the section on Taking Solr to
> Production.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
