[
https://issues.apache.org/jira/browse/COUCHDB-709?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12849569#action_12849569
]
Adam Kocoloski commented on COUCHDB-709:
----------------------------------------
Looks good to me.
The asynchronous shutdown problem is a symptom of us violating OTP design
principles in our structuring of CouchDB. The quartet of gen_servers
associated with each open database (couch_db, couch_db_updater,
couch_ref_counter, and couch_file) are not supervised, and as a result we have
a homegrown (async) system for killing them. The OTP supervisor system is
definitely synchronous.
> Restart actually restarts the server
> ------------------------------------
>
> Key: COUCHDB-709
> URL: https://issues.apache.org/jira/browse/COUCHDB-709
> Project: CouchDB
> Issue Type: Improvement
> Components: Database Core
> Reporter: Chris Anderson
>
> This patch will cause CouchDB to actually restart the server when a POST is
> made to /_restart
> The old way was unreliable as supervisors would shut things down
> asynchronously. This new way is much more brute force, which makes it more
> deterministic.
> This only really effects the test suite. I'm only pushing the patch through
> Jira to see if people see room for improvements.
> One improvement would be to add a timestamp for server boot time to the /
> response, but I seem to have avoided the need for that with my double GET
> magic.
> Do note: restart now drops any temporary config, hence the change to the
> reader_acls test.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
