lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Markus Jelsma <markus.jel...@openindex.io>
Subject RE: procedure to restart solrcloud, and config/collection consistency
Date Thu, 09 Feb 2017 13:12:25 GMT
Hello - see inline.
 
-----Original message-----
> From:xavier jmlucjav <jmlucjav@gmail.com>
> Sent: Thursday 9th February 2017 13:46
> To: solr-user <solr-user@lucene.apache.org>
> Subject: Re: procedure to restart solrcloud, and config/collection consistency
> 
> Hi Markus,
> 
> yes, of course I know (and use) the collections api to reload the config. I
> am asking what would happen in that scenario:
> - config updated (but collection not reloaded)
> - i restart one node
> 
> now one node has the new config and the rest the old one??

The restared/reloaded node has the new config, the others have the old config until reloaded/restarted.

> 
> Regarding restarting many hosts, my question is if we can just 'restart'
> each solr and that is enough, or it is better to first stop all, and then
> start all.

We prefer a rolling restart, restarting all nodes in sequence with some wait time in between
to allow the node to come back up properly. I see no reason to do a stop all/start all unless
you have cleared/will clear the index and want to reindex.

> 
> thanks
> 
> 
> On Thu, Feb 9, 2017 at 1:28 PM, Markus Jelsma <markus.jelsma@openindex.io>
> wrote:
> 
> > Hello - if you just want to use updated configuration, you can use Solr's
> > collection reload API call. For restarting we rely on remote provisioning
> > tools such as Salt, other managing tools can probably execute commands
> > remotely as well.
> >
> > If you operate more than just a very few machines, i'd really recommend
> > using these tools.
> >
> > Markus
> >
> >
> >
> > -----Original message-----
> > > From:xavier jmlucjav <jmlucjav@gmail.com>
> > > Sent: Thursday 9th February 2017 13:24
> > > To: solr-user <solr-user@lucene.apache.org>
> > > Subject: procedure to restart solrcloud, and config/collection
> > consistency
> > >
> > > Hi,
> > >
> > > When I need to restart a Solrcloud cluster, I always do this:
> > > - log in into host nb1, stop solr
> > > - log in into host nb2, stop solr
> > > -...
> > > - log in into host nbX, stop solr
> > > - verify all hosts did stop
> > > - in host nb1, start solr
> > > - in host nb12, start solr
> > > -...
> > >
> > > I always wondered, if this was not really needed, and I could just call
> > > 'restart' in every node, in a quick loop, and forget about it. Does
> > anyone
> > > know if this is the case?
> > >
> > > My doubt is in regards to changing some config, and then doing the above
> > > (just restart nodes in a loop). For example, what if I change a config G
> > > used in collection C, and I restart just one of the nodes (N1), leaving
> > the
> > > rest alone. If all the nodes contain a shard for C, what happens, N1 is
> > > using the new config and the rest are not? how is this handled?
> > >
> > > thanks
> > > xavier
> > >
> >
> 

Mime
View raw message