CoT should use the tasks scopes to query for private artifacts rather than its own
Categories
(Release Engineering :: Release Automation, enhancement)
Tracking
(Not tracked)
People
(Reporter: mtabara, Unassigned)
References
Details
Attachments
(1 file, 2 external files)
|
6.27 KB,
text/plain
|
Details |
In bug 1594776 we had a missing queue:get-artifact:releng/* scope in the TC client used to download the artifacts in CoT.
While talking to Tom about this, he suggested that the long term fix should be to allow scriptworker to use the tasks's permission to query for those private artifacts, as opposed to a dedicated client for it.
That said: we should enrich the release-partner signing jobs to encompass the queue:get-artifact:releng/* within their scopes so that CoT can re-use that permission, rather than its own.
Comment 1•7 years ago
|
||
We have should have those permissions in context.temp_queue. It looks like get_artifact_url always uses context.queue, which has the scriptworker scopes; we should allow it to use context.temp_queue if and when it exists, so we can pick up the task scopes.
Comment 2•7 years ago
|
||
We'll also need to adjust the in-tree code to provide that scope.
Comment 3•6 years ago
•
|
||
We're missing get-artifact scopes from the partner and eme-free scriptworker tasks. Once we add those, we can likely re-land the temp_queue fix. This isn't just the signing tasks, but beetmover as well.
Comment 5•6 years ago
|
||
This seems to do the right thing with ./mach taskgraph target-graph with a parameters.yml for a b8+. I haven't had a change to test it with taskgraph-diff, or in staging with dev scriptworkers using the temp queue credentials.
Comment 6•6 years ago
|
||
Tom wrote https://github.com/mozilla-releng/mvp-adhoc/compare/master...tomprince:artifacts#diff-6c5666bb90c294470f8f5f444221ae79R50 for fetch-artifact signing. I'm not 100% sure if it translates to non-fetch artifact signing.
Comment 7•6 years ago
|
||
We currently have the proxy enabled on docker-worker tasks (the actual repack, repackage), but not for anything using a scriptworker (beetmover, signing, notarization poller) or the chunking dummy on linux. Suspect no further changes required but testing will tell.
Comment 8•6 years ago
|
||
(In reply to Nick Thomas [:nthomas] (UTC+13) from comment #7)
We currently have the proxy enabled on docker-worker tasks (the actual repack, repackage), but not for anything using a scriptworker (beetmover, signing, notarization poller) or the chunking dummy on linux. Suspect no further changes required but testing will tell.
Correct. Scriptworker's context.temp_queue will essentially give us perms like the taskcluster proxy, at least where the queue is involved.
Comment 9•6 years ago
|
||
(In reply to Aki Sasaki [:aki] (he/him) (UTC-7) from comment #8)
(In reply to Nick Thomas [:nthomas] (UTC+13) from comment #7)
We currently have the proxy enabled on docker-worker tasks (the actual repack, repackage), but not for anything using a scriptworker (beetmover, signing, notarization poller) or the chunking dummy on linux. Suspect no further changes required but testing will tell.
Correct. Scriptworker's
context.temp_queuewill essentially give us perms like the taskcluster proxy, at least where the queue is involved.
Which, I'll note, is exactly how the taskcluster proxy is implemented. The worker gives the proxy the same credentials that we give to temp_queue, and it applies those credentials to calls through it, without exposing the credentials themselves to a task.
Comment 10•4 years ago
|
||
Comment 11•4 years ago
|
||
Dependent on https://github.com/mozilla-releng/scriptworker/issues/426
https://mozilla-hub.atlassian.net/browse/RELENG-917
Updated•3 years ago
|
Updated•1 year ago
|
Updated•1 year ago
|
Description
•