[
https://issues.apache.org/jira/browse/COUCHDB-3054?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15372922#comment-15372922
]
ASF subversion and git services commented on COUCHDB-3054:
----------------------------------------------------------
Commit 5e3bf1f5e2e96841ce0d2b26606c7f590c067bdc in couchdb-chttpd's branch
refs/heads/master from [~rnewson]
[ https://git-wip-us.apache.org/repos/asf?p=couchdb-chttpd.git;h=5e3bf1f ]
Tolerate failure of ensure_auth_ddoc_exists call
ensure_auth_ddoc_exists can crash if _users db does not exist, so move
it to the spawned listen_for_changes process where we already tolerate
the possibility of a crash (and have logic to restart after 5 seconds)
COUCHDB-3054
> chttpd cache invalidation is slow
> ---------------------------------
>
> Key: COUCHDB-3054
> URL: https://issues.apache.org/jira/browse/COUCHDB-3054
> Project: CouchDB
> Issue Type: Bug
> Reporter: Robert Newson
>
> The chttpd_auth_cache module follows the changes feed of the clustered _users
> database in order to invalidate the cache whenever a document changes.
> This can take over a minute to occur. It turns out we kill the changes feed
> process when we receive a heartbeat and restart the feed, leading to a long
> delay.
> We do want to stop periodically in order to close databases that might have
> been compacted.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
