Page MenuHomePhabricator

[spicerack] split SRE cookbooks into "shared" and "SRE-only"
Closed, DeclinedPublic

Description

At the moment, all SRE cookbooks are installed to both production (cuminXXXX) and cloud (cloudcuminXXXX) cumin hosts. I think it would be nice to separate the cookbooks that are relevant to both production and cloud hosts (e.g. sre.hosts.reimage) to a separate "shared" collection that can be installed on all cumin hosts, while the "SRE" collection could be installed only on production cumin hosts and removed from cloudcumins.

This could evolve even further if other teams/sub-teams want to create their own cookbook collection in the future. The modular structure introduced in Spicerack 7 makes it easy to cherry-pick a list of cookbook repos you want to install on a specific host.

Event Timeline

Restricted Application added a subscriber: Aklapper. · View Herald Transcript
fnegri triaged this task as Low priority.Aug 9 2023, 2:19 PM

That's indeed something we might want to look at going forward. The only blocker I see right now is that most of the "production" cookbook that could make sense to move to a shared base would not currently work from the cloudcumin hosts, making the point a bit moot.

True, sre.hosts.reimage is not likely to work anytime soon. The only sre.* cookbooks that I think we can easily run from cloudcumin are the sre.wikireplicas.* ones. But the ownership of wikireplicas is currently being discussed, so let's wait to see what is the outcome of that discussion.

I also wonder if we should remove production cookbooks from cloudcumin hosts, and keep only wmcs-cookbooks in the cookbooks_repos list. This means running cookbook -l in cloudcumin hosts would no longer return the list of sre.* cookbooks that do not work there.

Regarding keeping only the wmcs cookbooks in the conifg that's ok for me if it's ok for the WMCS team. At least for now, so we don't generate the expectation that those should work.

Change 955937 had a related patch set uploaded (by FNegri; author: FNegri):

[operations/puppet@production] [cluster::cloud_management] Don't install prod cookbooks

https://gerrit.wikimedia.org/r/955937

Change 955937 merged by FNegri:

[operations/puppet@production] [cluster::cloud_management] Don't install prod cookbooks

https://gerrit.wikimedia.org/r/955937

I merged the patch above and cleaned up the SRE cookbooks from cloudcumin[1-2]001.

I'm resolving this as "Declined" as there is no specific action we need now. I would open more specific tasks if we identify cookbooks that we want to share.