[jira] [Commented] (SOLR-6429) Unload core during tlog replay

["Erick Erickson (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-6429?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16962106#comment-16962106 ] 

Erick Erickson commented on SOLR-6429:
--------------------------------------

Can we close this?

> Unload core during tlog replay
> ------------------------------
>
>                 Key: SOLR-6429
>                 URL: https://issues.apache.org/jira/browse/SOLR-6429
>             Project: Solr
>          Issue Type: Bug
>          Components: SolrCloud
>    Affects Versions: 4.8.1
>            Reporter: Eran H
>            Priority: Major
>              Labels: core, exception, log, replay, timeout, transaction, unload
>
> Hi,
> I use solr 4.8.1 and solrj . When my server comes up it starts solr. If my server run configuration is to delete all data and start again fresh, I unload my solr core and then create it again. During this time, solr comes up.
> Solr receives the core unload request while replaying its transaction log. 
> Any logical behavior would be accepted in this point:
> 1. Exception that it cannot drop the core because it's in the middle of log replaying.
> 2. Wait until the log replay is finish and then drop the core.
> 3. FORCE to drop the core and stop the replay.
> 4. Provide API so I would know to question solr if it's "busy".
> 5. Etc.
> What really happens is the only illogical behavior:
> I get a timeout exception(!) but it continues to replay the logs. 
> core.properties file is deleted, core.properties.unloaded file is created,  and the folder is not deleted as I asked (using solrj). I can't delete the folder myself because it's locked, and I'm basically stuck with the core folder but with an unloaded core. I can't create the core again because the folder already exists, and I can't unload the core again because it does not exist!
> If solr receives a core unload request during tlog replay, it should either reject it with a dedicated exception (not timeout) or process it fully. Currently it tries to do both and it won't work.
> Thanks!



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org
For additional commands, e-mail: issues-help@lucene.apache.org