[
https://issues.apache.org/jira/browse/COUCHDB-2657?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14500885#comment-14500885
]
Russell Branca commented on COUCHDB-2657:
-----------------------------------------
I tracked this down to making a `fabric:open_doc` call from within the
`cassim_metadata_cache` gen_server. The problem is that `fabric:open_doc` does
a `receive` call while waiting for a worker response, and this intercepts any
other messages sent in. I've got a patch out [1] that shows the rough solution.
I did notice however that in the test case we're still getting a 409 error on
the second request, so we'll want to make sure we handle that properly. Should
we check and see if the values are the same? and if not reload? That might be
worthwhile but potentially expensive. Alternatively we could automatically
create the security doc on db creation.
Also, I think COUCHDB-2659 might be a duplicate of this issue.
[1] https://github.com/apache/couchdb-cassim/tree/2657-fix-cassim-fabric-calls
> if _metadata is there, creating a db does not work
> --------------------------------------------------
>
> Key: COUCHDB-2657
> URL: https://issues.apache.org/jira/browse/COUCHDB-2657
> Project: CouchDB
> Issue Type: Bug
> Security Level: public(Regular issues)
> Components: BigCouch, Database Core
> Reporter: Robert Kowalski
> Priority: Blocker
>
> with the new _setup_cluster endpoint a _metadata db is created.
> when it exists, and we create a db, read it _all_doc immediatly [1] (Fauxton
> does that) we get:
> `document update conflict`
> [1]
> http://localhost:8000/sdfsdtest/_all_docs?startkey=%22_design%2F%22&endkey=%22_design0%22&include_docs=true&limit=501
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
