summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
Paul J. Davis [Wed, 22 Feb 2017 20:21:09 +0000 (14:21 -0600)]
Merge remote-tracking branch 'cloudant/revert-default-value-for-max_write_delay'
brkolla [Mon, 20 Feb 2017 18:11:01 +0000 (13:11 -0500)]
Rollback change in default value for max_write_delay config
The default for `global_change:max_write_delay` config was changed from
500 to 25 and is causing performance regression. Reverting this change
back to 500 is resulting in improved performance.
COUCHDB-3304
ILYA Khlopotov [Fri, 2 Sep 2016 17:39:33 +0000 (10:39 -0700)]
Merge remote branch 'cloudant:3102-fix-return-value'
This closes #16
Signed-off-by: ILYA Khlopotov <iilyak@ca.ibm.com>
ILYA Khlopotov [Fri, 2 Sep 2016 17:19:36 +0000 (10:19 -0700)]
Return {ok, State} from handle_config_change/5
This is a fixup commit for COUCHDB-3102.
config_listener expects either remove_handler or {ok, State} from the
callback module.
COUCHDB-3102
Robert Newson [Fri, 2 Sep 2016 10:33:02 +0000 (11:33 +0100)]
use shard suffix when generating _changes ETag
In CouchDB 2.0, instance_start_time is always 0.
This means that when generating ETag values derived
from the database info object, the same ETags can be
incorrectly deemed valid between different database
instances with the metadata.
To avoid this we can incorporate the unique shard
suffix for the database instance when generating
the current ETag value.
Copied from fix by Will Holley in chttpd
COUCHDB-3017
ILYA Khlopotov [Tue, 23 Aug 2016 21:59:45 +0000 (14:59 -0700)]
Merge remote branch 'cloudant:3102-fix-config_subscription'
This closes #15
Signed-off-by: ILYA Khlopotov <iilyak@ca.ibm.com>
ILYA Khlopotov [Wed, 17 Aug 2016 18:55:59 +0000 (11:55 -0700)]
Update handle_config_terminate API
COUCHDB-3102
Jan Lehnardt [Sun, 17 Apr 2016 18:21:37 +0000 (20:21 +0200)]
add missing change_callback implementation
Jan Lehnardt [Sun, 17 Apr 2016 17:31:22 +0000 (19:31 +0200)]
enable feed=eventsource for /_db_updates
The fix is adding `eventsource` to the allowed options to
/_db_updates?feed=eventsource, since the underlying implementation
is a global database’s _changes feed.
Closes COUCHDB-2665
Alexander Shorin [Sun, 23 Aug 2015 10:32:37 +0000 (13:32 +0300)]
Integrate with Travis CI
Alexander Shorin [Thu, 1 Oct 2015 16:17:24 +0000 (19:17 +0300)]
Merge remote-tracking branch 'github/pr/13'
Nick Vatamaniuc [Thu, 1 Oct 2015 14:48:46 +0000 (10:48 -0400)]
Fix test.
Needs global_changes app started and get host from chttpd.
ILYA Khlopotov [Tue, 29 Sep 2015 20:05:05 +0000 (13:05 -0700)]
Pass supervisor's children to couch_epi
ILYA Khlopotov [Mon, 28 Sep 2015 16:48:59 +0000 (09:48 -0700)]
Update to new couch_epi API
ILYA Khlopotov [Wed, 2 Sep 2015 16:49:00 +0000 (09:49 -0700)]
Register service
COUCHDB-2796
ILYA Khlopotov [Tue, 1 Sep 2015 13:26:34 +0000 (06:26 -0700)]
Remove `ignore_providers` option
COUCHDB-2796
Alexander Shorin [Sun, 23 Aug 2015 10:31:47 +0000 (13:31 +0300)]
Add license header
Robert Newson [Wed, 12 Aug 2015 15:40:06 +0000 (16:40 +0100)]
Merge remote-tracking branch 'cloudant/epi_hooks'
ILYA Khlopotov [Fri, 31 Jul 2015 13:48:56 +0000 (06:48 -0700)]
Introduce `global_changes_plugin:transform_change/3`
Robert Newson [Wed, 29 Jul 2015 15:21:39 +0000 (16:21 +0100)]
Merge remote-tracking branch 'cloudant/chttpd-qs'
Robert Newson [Mon, 27 Jul 2015 12:13:11 +0000 (13:13 +0100)]
Handle waiting_for_updates in callback
COUCHDB-2724
ILYA Khlopotov [Thu, 23 Jul 2015 20:12:22 +0000 (13:12 -0700)]
Use `chttpd:qs*` instead of `couch_httpd:qs*`
ILYA Khlopotov [Wed, 15 Jul 2015 15:47:59 +0000 (08:47 -0700)]
Use dynamic handlers for `_db_updates`
Robert Newson [Thu, 2 Jul 2015 21:59:49 +0000 (22:59 +0100)]
erlang R14-18 compatibility
ILYA Khlopotov [Mon, 27 Apr 2015 14:41:50 +0000 (07:41 -0700)]
Fix _db_changes respone format
We are not suppose to return `account` field
`dnname` field is renamed to `db_name`
COUCHDB-2667
ILYA Khlopotov [Mon, 27 Apr 2015 14:41:29 +0000 (07:41 -0700)]
Fix indentation
Alexander Shorin [Sat, 18 Apr 2015 13:33:47 +0000 (16:33 +0300)]
Make global_changes database system one
Robert Newson [Sat, 11 Apr 2015 15:37:59 +0000 (16:37 +0100)]
handle heartbeat=false
Alexander Shorin [Fri, 27 Feb 2015 14:47:48 +0000 (17:47 +0300)]
Merge remote-tracking branch 'iilyak/2585-allowed_owner-hook'
This closes #3
ILYA Khlopotov [Tue, 24 Feb 2015 21:17:54 +0000 (13:17 -0800)]
Validate callback defined
COUCHDB-2585
ILYA Khlopotov [Wed, 18 Feb 2015 15:48:16 +0000 (07:48 -0800)]
Use `config` module for `allowed_owner` setting
COUCHDB-2585
ILYA Khlopotov [Mon, 16 Feb 2015 20:05:57 +0000 (12:05 -0800)]
Introduce an `allowed_owner` hook
COUCHDB-2585
Alexander Shorin [Wed, 4 Feb 2015 15:43:19 +0000 (18:43 +0300)]
Merge remote-tracking branch 'iilyak/2561-make-config-API-consistent'
This closes #2
COUCHDB-2561
ILYA Khlopotov [Fri, 30 Jan 2015 19:20:28 +0000 (11:20 -0800)]
Don't restart event handler on termination
COUCHDB-2561
ILYA Khlopotov [Thu, 29 Jan 2015 21:50:58 +0000 (13:50 -0800)]
Update config_listener behaviuor
COUCHDB-2561
Alexander Shorin [Mon, 26 Jan 2015 04:06:33 +0000 (07:06 +0300)]
Use ADMIN_CTX macro from couch_db.hrl
Robert Newson [Fri, 23 Jan 2015 16:33:33 +0000 (16:33 +0000)]
Open global changes db with admin context
The global changes feed should work even if the security object says otherwise.
BugzID: 43665
Alexander Shorin [Mon, 22 Dec 2014 15:14:18 +0000 (18:14 +0300)]
Fix LICENSE indention
Paul J. Davis [Fri, 22 Aug 2014 10:05:59 +0000 (05:05 -0500)]
Fix race condition by relying on admin only
Since we no longer have to worry about filtering for users we can let
fabric take care of the underlying limit implementation for now. Without
this there's a race condition where fabric will wait for Timeout seconds
before actually stopping. Any activity in that window will be sent
through the callback and forwarded on to clients. There's probably a way
to fix that but my brain isn't working so well right now.
Paul J. Davis [Thu, 21 Aug 2014 06:56:48 +0000 (01:56 -0500)]
Update to use couch_stats
Paul J. Davis [Wed, 13 Aug 2014 20:54:17 +0000 (15:54 -0500)]
Reimplement global_change rate limiting
Rather than just checking when a change occurs to update we instead just
update after max_write_delay milliseconds.
Robert Newson [Wed, 16 Jul 2014 12:06:58 +0000 (13:06 +0100)]
Add ASF license
robfraz [Thu, 20 Feb 2014 13:48:19 +0000 (13:48 +0000)]
Raise default global_changes max_write_delay & max_event_delay to 500ms
BugzID: 28257
Benjamin Bastian [Wed, 19 Feb 2014 21:41:28 +0000 (13:41 -0800)]
Use gen_server:cast rather than rexi RPCs
Prior to this commit, the global_changes_listener would send updates to
the global_changes_server via a rexi RPC which only did a
gen_server:call. This is needlessly heavyweight and would cause
gen_server:call timeout errors when the global_changes_server was slow
(usually due to slow disk IO). This commit removes the rexi RPC and
replaces it with a gen_server:cast.
Since this causes all global_changes_server handlers to not reply, this
commit also removes the format_reply function as it's unnecessary.
BugzID: 28242
Benjamin Bastian [Fri, 13 Dec 2013 22:31:30 +0000 (14:31 -0800)]
Make changes_callback end request when limit=1
Prior to this commit, the limit was hit when limit=0 by the first
changes_callback function clause. This makes maybe_finish end the
request when limit=1, so that it doesn't have to wait for another change
to end the request.
BugzID: 26166
Paul J. Davis [Fri, 13 Dec 2013 20:38:23 +0000 (14:38 -0600)]
This was a silly off-by-one error while counting
The start message doesn't count as a row so we don't want to adjust
Acc#acc.limit based on that message.
BugzId: 26165
Benjamin Bastian [Tue, 10 Dec 2013 21:50:00 +0000 (13:50 -0800)]
Fix limit=N for non-admins by counting post-filter
Prior to this commit, the _db_updates endpoint would often return less
than N results for limit=N queries, even when there were N changes to
return. This was because the limiting was done by passing the limit
parameter to fabric:changes, which doesn't account for the per-user
filtering done in global_changes_httpd.
This commit adds explicit counting of the filtered changes in
global_changes_httpd and manually ends the request when N post-filter
changes have been seen.
BugzID: 25272
Paul J. Davis [Thu, 12 Dec 2013 18:33:50 +0000 (12:33 -0600)]
Fix changes API usage for new pending values
BugzId: 26119
Robert Newson [Fri, 22 Nov 2013 16:54:32 +0000 (16:54 +0000)]
set module version to 1
Benjamin Bastian [Tue, 15 Oct 2013 18:50:07 +0000 (11:50 -0700)]
Fix heartbeat=true for db_updates feeds
This commit fixes some bad case clause ordering. It'd match 'true' to a
variable when the variable should only either be false or a number.
BugzID: 24170
Benjamin Bastian [Fri, 6 Sep 2013 01:33:38 +0000 (18:33 -0700)]
Add the initial version of the global_changes app.
BugzID: 17681
Benjamin Bastian [Fri, 9 Aug 2013 12:47:35 +0000 (13:47 +0100)]
Initial commit.