[
https://issues.apache.org/jira/browse/SOLR-7344?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14510045#comment-14510045
]
Yonik Seeley commented on SOLR-7344:
------------------------------------
(regarding forwarded requests, where a request received by a node w/o a core in
the target collection will forward to a node with a core in the target
collection)
bq. The original request would block on a thread in the external endpoint and
the forwarded request would be executed on a thread belonging to internal
thread-pool. So there is no chance of deadlock as such.
That essentially means that top-level requests could go to the "internal"
request pool and we are back to square 1: threads from that request pool (the
internal one now) depending on sub-requests that use the same pool. That leads
to distributed deadlock if the pool is limited. It's easier to think about
these things if you just take it to the extremes... think of pool sizes of 1.
> Use two thread pools, one for internal requests and one for external, to
> avoid distributed deadlock and decrease the number of threads that need to be
> created.
> ---------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: SOLR-7344
> URL: https://issues.apache.org/jira/browse/SOLR-7344
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Reporter: Mark Miller
> Attachments: SOLR-7344.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
