[jira] [Commented] (SOLR-6846) deadlock in UninvertedField#getUninvertedField()

["Prasanna Ranganathan (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-6846?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15158240#comment-15158240 ] 

Prasanna Ranganathan commented on SOLR-6846:
--------------------------------------------

I was looking into heap/memory leak issues on our Solr nodes and found the same symptom with many threads stuck on cache.wait() call when analyzing the heap dump. I think the issue is caused by the following:

{quote}
uif = new UnInvertedField(field, searcher);
{quote}

The above constructor can throw an IOException. If there happens to be any exception thrown in the constructor then there will be a bunch of threads stuck on the cache.wait() call. Hence the exception needs to be caught and re-thrown after clearing the placeholder object and notifying all the waiting threads.

> deadlock in UninvertedField#getUninvertedField()
> ------------------------------------------------
>
>                 Key: SOLR-6846
>                 URL: https://issues.apache.org/jira/browse/SOLR-6846
>             Project: Solr
>          Issue Type: Bug
>          Components: search
>    Affects Versions: 4.10.2
>            Reporter: Avishai Ish-Shalom
>             Fix For: 5.0, master
>
>         Attachments: SOLR-6846.patch
>
>
> Multiple concurrent calls to UninvertedField#getUninvertedField may deadlock: 
> if a call gets to {{cache.wait()}} before another thread gets to the synchronized block around {{cache.notifyAll()}} code will deadlock because {{cache.wait()}} is synchronized with the same monitor object.



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

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