1172023 - deploy funsize scheduler in releng network

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[[e-mail] (:bhearsum)]

Probably a few Buildbot Masters.

Comment 1

[[e-mail] (:bhearsum)]

Attachment: add puppet manifests for funsize scheduler

Nothing too fancy here, the only special part was the need to add package for libffi, because we need libffi-devel to install the cffi python module.

Comment 2

[[e-mail] (:bhearsum)]

Attachment: fix config file typo + add root cert location

Very little in this interdiff, just a couple of issues I discovered while testing on a dev machine.

Comment 3

[Dustin J. Mitchell [:dustin] (he/him)]

Comment on attachment 8616884 [details] [diff] [review]
fix config file typo + add root cert location

Review of attachment 8616884 [details] [diff] [review]:
-----------------------------------------------------------------

I can't see the interdiff, but I'll take your word for it..

Comment 4

[Rail Aliiev [:rail]]

BTW, I just landed a cahnge which replaces pystache with Jinja2. Shouldn't be hard to replace.

Comment 5

[[e-mail] (:bhearsum)]

Comment on attachment 8616884 [details] [diff] [review]
fix config file typo + add root cert location

I've landed this and it will be deployed on bm89, 91, and 94 soon. It won't actually work until we get the Balrog accounts sorted out in bug 1172026, but it won't cause any harm.

Comment 6

[Rail Aliiev [:rail]]

Attachment: jinja-puppet.diff

* switch to Jinja2
* deploy newer version

Comment 7

[[e-mail] (:bhearsum)]

I started up on of the instances by hand and discovered that we'll need netflows for aus4-admin.mozilla.org:443 added before these will work, right now the connections are timing out. We're also waiting on bug 1172026 so we have isolated accounts for Funsize.

Comment 8

[[e-mail] (:bhearsum)]

Comment on attachment 8617901 [details] [diff] [review]
jinja-puppet.diff

Review of attachment 8617901 [details] [diff] [review]:
-----------------------------------------------------------------

I see you've already got the new packages in place, wunderbar!

Comment 9

[Rail Aliiev [:rail]]

Comment on attachment 8617901 [details] [diff] [review]
jinja-puppet.diff

remote:   https://hg.mozilla.org/build/puppet/rev/82d16f33a1b8
remote:   https://hg.mozilla.org/build/puppet/rev/70fbeb983947

Comment 10

[[e-mail] (:bhearsum)]

So, the service itself is deployed and capable of running now. We're still waiting on bug 1172026 in order to be able to talk to Balrog successfully, though.

Rail, is there any testing you want to do against balrog dev in the meantime? Or can we just enable the scheduler against Balrog prod when we the new credentials?

Comment 11

[[e-mail] (:bhearsum)]

Rail told me in person that we're fine to start up the funsize scheduler against production. I've granted "funsize" PUT permissions to /releases/:name/builds/:platform/:locale for all products on Balrog prod, and updated Hiera with the real LDAP password.

Rail says there's a code change that we need before things will work, so we're waiting on a patch for that.

Comment 12

[Rail Aliiev [:rail]]

Dev testing can be done in parallel, because we don't share the queues.

Comment 13

[Rail Aliiev [:rail]]

Attachment: auto-re-start.diff

Comment 14

[Rail Aliiev [:rail]]

Comment on attachment 8624283 [details] [diff] [review]
auto-re-start.diff

remote:   https://hg.mozilla.org/build/puppet/rev/e60bb8d6294f
remote:   https://hg.mozilla.org/build/puppet/rev/3e6db55bc911