1765658 - [meta] allow for testing fxci tasks on the staging cluster

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Aki Sasaki (not active)]

We had this, but we've regressed.
Meta bug to track the fixes needed.

Comment 1

[Aki Sasaki (not active)]

1. run remove_secrets_from_staging.py to clear out the previous secrets I populated (with fake or non-sensitive values)
2. graft this patch onto central, push to try via ./mach try release --migration central-to-beta -v 102.0b1
3. we don't need the production try tasks, so cancel the graph via taskcluster group cancel -- TASK_GROUP_ID or the cancel-all action to save compute cycles. Use tc-relduty for the former, or sign into Treeherder or TC UI for the latter.
4. copy the try revision; you'll need this for the next step
5. Use fxci to send that task to the staging cluster:

# Set the revision to your try push revision
REVISION=12345
tc-staging  # using the aliases in https://docs.mozilla-releng.net/en/latest/taskcluster/taskcluster_cli.html#aliases to set your TASKCLUSTER_ROOT_URL to staging
eval $(taskcluster signin --expires 15m)
fxci replay-hg-push try $REVISION
# This will output a url like `https://stage.taskcluster.nonprod.cloudops.mozgcp.net/tasks/J9WeztDYT4aQstuJUGOgIg`.

6. This is an on-push task that will trigger a decision task; look at the logs to find the decision task url/taskId. This will look like:

    ],
    "schedulerId": "gecko-level-1",
    "scopes": [
        "assume:repo:hg.mozilla.org/try:branch:default",
        "queue:route:[e-mail].*",
        "in-tree:hook-action:project-gecko/in-tree-action-1-*",
        "index:insert-task:gecko.v2.try.*"
    ],
    "tags": {
        "createdForUser": "[e-mail]",
        "kind": "decision-task"
    },
    "taskGroupId": "a91geUF6Tlm7hT1aB41CXA",
    "workerType": "decision"
}
Task Id: a91geUF6Tlm7hT1aB41CXA
[taskcluster 2022-05-10 20:14:34.072Z] === Task Finished ===
[taskcluster 2022-05-10 20:14:34.073Z] Successful task run with exit code: 0 completed in 14.569 seconds

In the above case, we want a91geUF6Tlm7hT1aB41CXA; go to https://stage.taskcluster.nonprod.cloudops.mozgcp.net/tasks/a91geUF6Tlm7hT1aB41CXA in that case.

This will schedule a level 1 staging taskGroup for a staging release. We expect:

• ideally a green decision task
• all scriptworker tasks will hang and hit the exception deadline-exceeded at some point, because there are no scriptworkers pointed at the staging cluster to claim those tasks
• same with the hardware pools

We at least want to see a good number of docker-worker tasks going green.
This is happening now, so I'll go ahead and get the above into the docs.

Comment 2

[Aki Sasaki (not active)]

Attachment: docs PR

Comment 3

[Aki Sasaki (not active)]

This should be done once we land the above PR.