Every once in a while we need to cancel a task-graph for some reason, and along with it all of its dependencies. That might include fail fast scenarions in which, for example, various credentials are not longer good, etc. In order to do that there are two specific pieces of information that you need in hand:
- the task-graph-id
- full tctalker setup
For a few weeks now, this information lies within the email that's being sent by Ship-it. So just grab it from there:
....
A new build has been submitted through ship-it:
...
Task graph: https://tools.taskcluster.net/task-graph-inspector/#<task-graph-id>/
...
- navigate to Taskcluster tools and make sure you're logged-in
- make sure the following scopes are listed under your Taskcluster credentials:
auth:create-client:mozilla-ldap/<ldap_username>/*
auth:create-client:project/releng/*
- navigate to Taskcluster clients and create a new clientId - dedicated for playing with tctalker.
- it should already be pre-formatted to
mozilla-ldap/<ldap_username>/
at which you can concatenate any string. You can simply add tctalker. - use queue:* for Client Scopes
- hit the Create Client button and make sure you save the Access token that is generated
- create a json config file (e.g. "config.json") on disk that looks like this:
{
"credentials": {
"clientId": "you-will-never-guess",
"accessToken": "nor here!"
}
}
- clone yourself a copy of tctalker and change directory to the clone repo
- run the following:
python src/tctalker/tctalker.py --conf config.json cancel_graph <task-graph-id>