1269339 - Update generic worker version used in worker type win2012 to v2.0.0alpha31

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Pete Moore [:pmoore][:pete]]

Attachment: Github Pull Request for OpenCloudConfig

There have been some changes to the generic worker which should be applied to the RelOps win2012 worker type. Most notably, bug 1267992 and bug 1268945 have been fixed in generic-worker release v2.0.0alpha31.

I've submitted a PR for this based on my understanding of how it should work. I may be wildly off though, so please check it carefully! :)

Comment 1

[Rob Thijssen [:grenade (EET/UTC+0300)]]

Comment on attachment 8747693 [details] [review]
Github Pull Request for OpenCloudConfig

merged

Comment 2

[Rob Thijssen [:grenade (EET/UTC+0300)]]

on upgrade to alpha 32 something strange happens to the GenericWorker user account making it impossible for this user to log in (either via rdp or the autologin needed by g-w.exe) The error message states that the user profile is missing or corrupt.
Deleting the account and recreating it has no effect (the new account also has profile issues).
The install of the worker seems to have gone well during ami creation but the worker does not start when new instances are instantiated due to problems with the autologin and the account.

Comment 3

[Rob Thijssen [:grenade (EET/UTC+0300)]]

running the worker manually (by executing c:\generic-worker\run-generic-worker.bat in a cmd prompt on the worker instances) results in tasks starting but throwing exceptions immediately:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=048eec7365f00c8f03f0675a47eb7d958ae264cc&selectedJob=20414921

Comment 4

[Pete Moore [:pmoore][:pete]]

Unfortunately I'm not able to reproduce on the AMI I created. Would you be able to grab a generic-worker.log file from a failing machine?

If the login isn't working for that user, I would expect the tasks to fail, but I'm not sure why the user has not been created successfully. I wonder if there is a race condition for the user being created, and the machine shutting down to be snapshotted. Perhaps the commands to create the user return immediately, before they have completed, when running generic-worker install startup? Do you know if the generic worker install is (one of) the last thing(s) done before the machine shuts down? I'm not sure this would be the cause, but could be worth investigating.

Comment 5

[Rob Thijssen [:grenade (EET/UTC+0300)]]

doesn't happen on new gw2 alphas