Hi Erick,

Your comment does not reflect on the Jira.

I also updated the MERGEINDEXES documentation (
https://cwiki.apache.org/confluence/display/solr/Merging+Indexes ) to
reflect that the WAR is pre-extracted from Solr 5.3 onwards.

On Thu, Sep 10, 2015 at 5:39 AM, Erick Erickson <[email protected]>
wrote:

> Hmmm...... It'd be great to have this documented!
>
> I gave this a quick shot just to see if it'd do what I'd expect and
> it's not actually that hard:
>
> 0> created the "techproducts" non-solr-cloud collection
> 1> shut down Solr
> 2> moved the entire directory "somewhere else", not in the Solr tree
> to simulate, say, bringing it over from some other machine.
> 3> brought up ZK and pushed the configuration file up
> 4> started SolrCloud (nothing is in it as you'd expect)
> 5> created a new collection with the config from step <3> (name irrelevant)
> 6> shut down the cloud
> 7> Copied just the _contents_ of the index directory from step <0> to
> the index directory created in <5>
> 8> restarted SolrCloud
>
> And all was well.
>
> I also tried just creating a new collection (1 shard) and using
> MERGEINDEXES with the indexDir option which also worked. I think I
> like that a little better, there are fewer places to mess things up,
> and it doesn't require bouncing SolrCloud. The first time I tried it I
> didn't manage to issue the commit, so that should be called out. Also
> should call out checking that the doc count is right since if a person
> gets nervous and issues the merge N times you have Nx the docs...
>
> You'd want ADDREPLICAs once you were satisfied you'd moved the index
> correctly of course. And hope that the config you pushed up was
> actually OK. Perhaps something here about just moving the relevant
> parts of schema.xml rather than the whole (old) config dir? Or maybe
> even proofing things out on 5x first?
>
> Of course, all this assuming you couldn't just re-index fresh ;).
>
> FWIW,
> Erick
>
>
>
> On Wed, Sep 9, 2015 at 4:31 PM, Shawn Heisey (JIRA) <[email protected]>
> wrote:
> >
> >     [
> https://issues.apache.org/jira/browse/SOLR-8027?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14737799#comment-14737799
> ]
> >
> > Shawn Heisey commented on SOLR-8027:
> > ------------------------------------
> >
> > I can try out some things tonight when I get home, assuming the honeydew
> list is not extreme.
> >
> >> Reference guide instructions for converting an existing install to
> SolrCloud
> >>
> ----------------------------------------------------------------------------
> >>
> >>                 Key: SOLR-8027
> >>                 URL: https://issues.apache.org/jira/browse/SOLR-8027
> >>             Project: Solr
> >>          Issue Type: Improvement
> >>          Components: documentation
> >>            Reporter: Shawn Heisey
> >>
> >> I have absolutely no idea where to begin with this, but it's a definite
> hole in our documentation.  I'd like to have some instructions that will
> help somebody convert a non-cloud install to SolrCloud.  Ideally they would
> start with a typical directory structure with one or more cores and end
> with cores named foo_shardN_replicaM.
> >> As far as I know, Solr doesn't actually let non-cloud cores coexist
> with cloud cores.  I once tried to create a non-cloud core on a cloud
> install, and couldn't do it.
> >
> >
> >
> > --
> > This message was sent by Atlassian JIRA
> > (v6.3.4#6332)
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: [email protected]
> > For additional commands, e-mail: [email protected]
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>


-- 


Regards,
Varun Thacker

Reply via email to