[
https://issues.apache.org/jira/browse/SOLR-6392?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15035222#comment-15035222
]
Erick Erickson commented on SOLR-6392:
--------------------------------------
Frank:
Here's the rules if you don't specify a configName when issuing the Collections
API CREATE command (at least as I understand them)
0> if there are no config sets in ZK, fail
1> if there's a configset with the same name as the collection use it.
2> else if there's only a single configset, use that regardless of whether name
matches or not.
3> else if there's more than one configset, fail.
The above if (and only if) there is no config specified when creating the
collection. If you do specify a configset name, then of course it has to be
there or collection creation fails.
So here's one possible sequence leading to what's described here:
1> upload a config set with a name of "bonkers".
2> create collection1 with or without specifying the configset name
3> create collection2 with or without specifying the configset name
Now, both are registered with ZK as using configset "bonkers" and will stay
that way unless and until that association is explicitly changed via linkconfig
as Mark recommended.
It would be dangerous IMO to change the config set associated with collection2
just because someone uploaded a new configset with the name collection2.
So I believe this is working as designed, I'm going to close it. We can reopen
or create a new JIRA if necessary.
> If run Solr having two collections configured but only one config delivered
> to Zookeeper causes that config is applied for all collections
> ------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: SOLR-6392
> URL: https://issues.apache.org/jira/browse/SOLR-6392
> Project: Solr
> Issue Type: Bug
> Affects Versions: 4.4
> Reporter: Ilya Meleshkov
>
> I have simplest Solr cloud configured locally. Thus I have single Solr and
> Zookeeper nodes.
> Steps to reproduce an error:
> # have stopped Solr+ZK with two collections
> # run ZK
> # deliver config to one collection only
> # run Solr - Solr running without any complains or errors
> # deliver config to second collection - doesn't have an effect
> But if I deliver configs for both collections before start Solr - it work
> perfectly.
> So I would say that Solr should fail with meaningful error if there is no
> config for some collection.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
