Ed Morley [:emorley] Reporter
	Description • 10 years ago

[emorley@treeherderadm.private.scl3 treeherder-service]$ cat deployment/update/commander_settings.py | grep 'HOSTGROUP'

WEB_HOSTGROUP = 'treeherder-web'
CELERY_HOSTGROUP = 'treeherder-processors'


These hostgroups are used by the deploy script:
https://github.com/mozilla/treeherder-service/blob/f7fed3e46ebebc2dfc7dabf331057dbb48db8d7d/deployment/update/update.py#L65

The new treeherder etl nodes are missing - so they won't get updated on Chief deploy.

	Ed Morley [:emorley] Reporter
	Comment 1 • 10 years ago

Once prod's version is updated, I'll update the example file in the repo:
https://github.com/mozilla/treeherder-service/blob/master/deployment/update/commander_settings.py-dist

	Ed Morley [:emorley] Reporter
	Comment 2 • 10 years ago

Also, perhaps we should just check commander_settings.py into the repo, and not have a split example-real file workflow? There's nothing secret in the prod file that can't be checked into the repo.

	Ed Morley [:emorley] Reporter
	Comment 3 • 10 years ago

Just remembered prod vs stage will need different values for DEPLOY_SCRIPT, so comment 2 won't work.

(Bah the delights of things that occur to you after pressing submit)

	Kendall Libby [:fubar] (he/him) Assignee
	Comment 4 • 10 years ago

for whatever reason, that file isn't in puppet. I've updated it on disk to be

CELERY_HOSTGROUP = ['treeherder-processors',
                    'treeherder-etl'
                   ]

we may have yet more bumps when it comes to staging, too, but I won't know for certain until I start refactoring things and trying to puppetize them.

	Ed Morley [:emorley] Reporter
	Comment 5 • 10 years ago

Thank you - I'll update the in-repo example file in another bug.

	Kendall Libby [:fubar] (he/him) Assignee
	Comment 6 • 10 years ago

hostgroups for were changed again in bug 1112290 to support restarting workers on each group:

WEB_HOSTGROUP = 'treeherder-stage-web'
ETL_HOSTGROUP = 'treeherder-stage-etl'
LOG_HOSTGROUP = 'treeherder-stage-processors'
RABBIT_HOSTGROUP = 'treeherder-stage-rabbit'

added commander_settings.py to both new and old puppet modules; old version still contains the old CELERY_HOSTGROUP (which is replaced by the three new ones above in staging).

	Ed Morley [:emorley] Reporter
	Comment 7 • 9 years ago

(In reply to Kendall Libby [:fubar] from comment #6)
> hostgroups for were changed again in bug 1112290 to support restarting
> workers on each group:
> 
> WEB_HOSTGROUP = 'treeherder-stage-web'
> ETL_HOSTGROUP = 'treeherder-stage-etl'
> LOG_HOSTGROUP = 'treeherder-stage-processors'
> RABBIT_HOSTGROUP = 'treeherder-stage-rabbit'
> 
> added commander_settings.py to both new and old puppet modules; old version
> still contains the old CELERY_HOSTGROUP (which is replaced by the three new
> ones above in staging).

[emorley@treeherderadm.private.scl3 treeherder-service]$ cat deployment/update/commander_settings.py
...
WEB_HOSTGROUP = 'treeherder-web'
ETL_HOSTGROUP = 'treeherder-etl'
LOG_HOSTGROUP = 'treeherder-processors'
RABBIT_HOSTGROUP = 'treeherder-rabbit'
CELERY_HOSTGROUP = ['treeherder-processors',
                    'treeherder-etl',
                    'treeherder-rabbitmq',
                   ]
...

Seems like CELERY_HOSTGROUP is now unused and so can be removed from that file?
Once we've finalised the prod file I'll update the in-repo example accordingly :-)

	Kendall Libby [:fubar] (he/him) Assignee
	Comment 8 • 9 years ago

Yup, it's unused. Removed the CELERY_HOSTGROUP.

	Ed Morley [:emorley] Reporter
	Comment 9 • 9 years ago

Thanks :-)