[ 
https://issues.apache.org/jira/browse/SOLR-7572?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14559960#comment-14559960
 ] 

Erick Erickson commented on SOLR-7572:
--------------------------------------

I've written a test program that hammers a cluster with updates and I'm not 
able to reproduce this yet so I'm not quite sure how to make progress on these. 
I'll be tweaking/running the test harness for a while to see what I can see.

> hard commits with waitSearcher=true occasionally returns without waiting 
> leading to inconsistent views of the index.
> --------------------------------------------------------------------------------------------------------------------
>
>                 Key: SOLR-7572
>                 URL: https://issues.apache.org/jira/browse/SOLR-7572
>             Project: Solr
>          Issue Type: Bug
>    Affects Versions: 4.10.3
>            Reporter: Erick Erickson
>            Assignee: Erick Erickson
>
> Another observed behavior while working with a customer, possibly related to 
> SOLR-7571. Again, assigning to myself but feel free to take it.
> Issuing a commit (openSearcher=true) sometimes returns immediately and does 
> not actually work as expected.
> Here's the process. 
> > Index a bunch of documents with timestamp=T1. Some time later update many 
> > (but not all) of those documents with a more recent timestamp=T2.
> > issue a hard commit waitSearcher=true
> > Issue a query ...fq=timestamp:[* TO T1]
> The return count includes many (all?) records with timestamp=T1. Simply 
> waiting for "a while" and re-issuing the query will eventually return the 
> correct number of docs.
> autocommits are disabled here, i.e. hard commits are happening with 
> openSearcher=false and soft commits are disabled in solrconfig.xml. There are 
> no other clients sending commit messages.
> Of course this is not reproducible at will and whether it's related to the 
> client firehosing updates to Solr I don't know yet.
> To pursue this I'm hoping to set up a "firehose" test scenario and see how 
> many of these I can reproduce.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to