[
https://issues.apache.org/jira/browse/COUCHDB-3095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15409590#comment-15409590
]
Eric Avdey commented on COUCHDB-3095:
-------------------------------------
Requests in waiting list of process manager getting assigned first returned
idle process without initializing it with teach_ddoc. With a chances this leads
to couchjs throwing "query_protocol_error" because it doesn't have requestor's
ddoc in its internal cache.
> Cluster starts throwing "query_protocol_error" once os process hard limit
> reached
> ---------------------------------------------------------------------------------
>
> Key: COUCHDB-3095
> URL: https://issues.apache.org/jira/browse/COUCHDB-3095
> Project: CouchDB
> Issue Type: Bug
> Reporter: Eric Avdey
> Assignee: Eric Avdey
>
> When a number of os processes reaches the hard limit a cluster starts to
> throw query_protocol_error errors.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
