Disable hypervisors by default because they don't have notifications. When enabled, put warning in indexing message.

Registered by Travis Tripp

Disable hypervisors by default because they don't have notifications. When enabled, put warning in indexing message.

Add documentation

15:15:26 <TravT> #topic Disable plugins by default that don't have notifications
15:15:38 <TravT> there may be an even bigger question here
15:15:51 <TravT> but basically, right now we have enabled most plugins by default
15:15:55 <TravT> swift being the exception
15:16:19 <TravT> we didn't enable swift by default because there aren't supported notifications yet.
15:17:15 <TravT> so, i was thinking that perhaps we should make it a policy that if a resource type has to be re-indexed periodically (via cron or some other scheduler), that we should disable them by default
15:17:36 <sjmc7> TravT: do you mean by default the example devstack conf? or in code?
15:17:40 <TravT> in code
15:18:02 <TravT> on devstack, enable by default in sample local.conf
15:18:43 <yingjun> It’s ok for me, just make sure document that..
15:19:25 <sjmc7> to go one step further, we could disable everything by default and make things opt-in
15:19:32 <TravT> yes, that is the next consideration
15:19:43 <rosmaita> i understand the motivation, but do you really think it's necessary?
15:19:59 <RickA-HP> Is there any way to make the connection of enabling a plugin and periodically re-indexing more explicit?
15:20:10 <sjmc7> i think delpoyers are probably going to write their own config file anyway, so the defaults don’t matter too much
15:20:17 <rosmaita> for a young project, probably don't want the devstack and shipping config to differ
15:20:23 <rosmaita> sjmc7: ++
15:20:39 <TravT> so, you can see my long winded comments on https://review.openstack.org/#/c/332242/
15:20:42 <TravT> patch set 4
15:21:11 <TravT> The last line kind of summarizes it
15:21:14 <TravT> All of the above has led me to also realize that upgrade scenarios may be problematic for deployers, particularly because we seem to just enable plugins by default or not. If you enable them by default, a deployer who had a config file per service will suddenly have to update every single config file to disable the new plugin (At least according to documentation:
15:21:14 <TravT> http://docs.openstack.org/developer/searchlight/plugins.html#non-inheritable-common-configuration-options)
15:22:05 <sjmc7> from my experience we don’t typically blindly re-use config files from release to release
15:22:27 <sjmc7> since so much changes with the various oslo libraries and in services themselves
15:24:07 <TravT> okay, so, sjmc7 rosmaita are you voting to leave all plugins enabled by default?
15:24:25 <TravT> or disable all or disable the ones without notifications by default?
15:24:26 * rosmaita is still reading that monster comment
15:24:41 <sjmc7> i think considering some as ‘experimental’ and disabling them by default is ok
15:24:58 <sjmc7> fully functional ones i’m ok leaving enabled
15:25:52 <TravT> yeah, my only reasoning on disable by default those which don't have notifications is that i wouldn't want a deployer to falsely think that they can expect the data to stay up to date in the index
15:26:28 <TravT> at least without them doing an external sync process
15:26:36 <david-lyle> perhaps a different section in the config file to point out the need for the cron job on those
15:26:50 <david-lyle> separate plugins that have notifications and those that don't
15:27:49 * TravT would have to think about that a bit...
15:28:12 <TravT> maybe if nothing else we could output a warning message in searchlight-manage index sync
15:28:21 <TravT> for any plugins that don't have notification listeneres?
15:28:33 <TravT> but are enabled
15:30:05 <TravT> okay, i think at least i've introduced the topic
15:30:05 <lei-zh> in index sync or searchlight-listener
15:30:28 <david-lyle> combination of warnings and clear documentation
15:30:37 <rosmaita> david-lyle: ++
15:30:44 <TravT> that sounds like a winner
15:31:18 <TravT> Okay, we'll go that direction
15:31:19 <TravT> thanks

Blueprint information

Travis Tripp
Travis Tripp
Series goal:
Accepted for newton
Milestone target:
milestone icon newton-3
Started by
Travis Tripp
Completed by
Travis Tripp

Related branches



Gerrit topic: https://review.openstack.org/#q,topic:bp/warn-for-types-without-notifications,n,z

Addressed by: https://review.openstack.org/349115
    Add warning for plugins without notifications


Work Items

This blueprint contains Public information 
Everyone can see this information.


No subscribers.