1600677 - gecko-t-linux-talos workers do not have `TASKCLUSTER_WORKER_LOCATION` environment variable set

Status: RESOLVED FIXED

(Infrastructure & Operations :: RelOps: General, defect)

Description

[Connor Sheehan [:sheehan]]

I've noticed in a few logs (1, 2, 3) that the TASKCLUSTER_WORKER_LOCATION environment variable isn't being set on releng-hardware/gecko-t-linux-talos workers. This variable is required to for run-task to find an appropriate private hgweb mirror to clone/pull from.

Comment 1

[:dhouse]

The env var, TASKCLUSTER_WORKER_LOCATION, gets created by generic-worker from the workerLocation variable:
https://github.com/taskcluster/generic-worker/blob/f4b6a7160e9d8d94d4655b9493737654fb8fec75/README.md#set-up-your-env
There are examples for aws and gcp. I'll have to look around more to find what it needs to be set to for the hardware workers.

Comment 2

[Connor Sheehan [:sheehan]]

Oh, these are non-cloud workers. In that case the workers should be cloning from the public hg.mozilla.org service anyways as we don't have mirrors for Mozilla datacentres. So this is still a bug, but it won't fix the problem I described in comment 0.

I'll leave this open for now since the variable should be set according to Taskcluster standards, but it's probably rather low priority.

Comment 3

[:dhouse]

(In reply to Connor Sheehan [:sheehan] from comment #2)

Oh, these are non-cloud workers. In that case the workers should be cloning from the public hg.mozilla.org service anyways as we don't have mirrors for Mozilla datacentres. So this is still a bug, but it won't fix the problem I described in comment 0.

I'll leave this open for now since the variable should be set according to Taskcluster standards, but it's probably rather low priority.

Thanks. That makes sense.

I'll keep this bug for adding the variable on the non-cloud linux (maybe others too):
it was added elsewhere through Bug 1572236
notes/proposal in:
https://github.com/taskcluster/taskcluster-rfcs/blob/master/rfcs/0148-taskcluster-worker-location.md