coordination #117673
open
coordination #154777: [saga][epic] Shareable os-autoinst and test distribution plugins
coordination #108527: [epic] os-autoinst wheels for scalable code reuse of helper functions and segmented test distributions
[epic][tools] sporadic "Unable to clone Git repository" for wheels
Added by szarate about 2 years ago.
Updated almost 2 years ago.
Category:
Feature requests
Estimated time:
(Total: 0.00 h)
Description
Observation¶
Every so often, the plugins fail to clone a repository for instance #117622#note-
https://openqa.suse.de/tests/9668977#
[2022-10-06T02:01:46.761898+02:00] [info] ::: OpenQA::Isotovideo::Utils::checkout_git_repo_and_branch: Cloning git URL 'https://github.com/Zaoliang/functional_wheel'
[2022-10-06T02:01:52.973889+02:00] [debug] Cloning into 'functional_wheel'...
fatal: unable to access 'https://github.com/Zaoliang/functional_wheel/': OpenSSL SSL_connect: Connection reset by peer in connection to github.com:443
Acceptance criteria¶
- AC1: A Plugin can be loaded on demand (maybe requires a bit more thinking, and a bit more of thought into the design)
- AC2: A GH token can be used for authenticated requests (I suspect that we're hitting the rate limit here)
- AC3: Wheels cloning is retried (A maximum retry of N times can be configured in the wheels.yaml)
- Copied from action #117622: [qe-core] Unable to clone Git repository for wheels added
- Subject changed from [tools] Unable to clone Git repository for wheels to [tools] sporadic "Unable to clone Git repository" for wheels
I think implementing 2 and 3 should be a good enough solution for starters
- Category changed from Regressions/Crashes to Feature requests
- Category changed from Feature requests to Regressions/Crashes
@okurz: We're reporting here that this new thing is randomly breaking unrelated tests. In what way is that a "feature request"?
- Category changed from Regressions/Crashes to Feature requests
MDoucha wrote:
@okurz: We're reporting here that this new thing is randomly breaking unrelated tests. In what way is that a "feature request"?
This isn't a regression, hence the category Feature requests.
cdywan wrote:
MDoucha wrote:
@okurz: We're reporting here that this new thing is randomly breaking unrelated tests. In what way is that a "feature request"?
This isn't a regression, hence the category Feature requests.
I'd argue it's a bug in the implementation, AC1 and AC2 are more of feature requests, than AC3 (which is a workaround/solution to the bug)... but semantics is a different thing, for now the possibility of having 403 HTTP Errors, blocks anybody from being able to properly use the plugin system for the test distribution.
Keep in mind that the category is based on what's known to work in released code, not what is desirable or common sense.
How often does this occur? Maybe the prio should be raised? Worst case the wheels.yaml could be dropped temporarily to avoid affecting many tests.
cdywan wrote:
How often does this occur? Maybe the prio should be raised? Worst case the wheels.yaml could be dropped temporarily to avoid affecting many tests.
It happens each time GitHub has an outage or OpenQA workers lose network access to the outside world. If GitHub goes down for 10 minutes while the OpenQA queue is full of livepatch tests, we'll end up with 8000+ failed jobs because of that.
[…] for now the possibility of having 403 HTTP Errors, blocks anybody from being able to properly use the plugin system for the test distribution.
I agree. I suggest to actually disable the use of wheels in os-autoinst-distri-opensuse until this has been resolved.
- Parent task set to #108527
- Tracker changed from action to coordination
- Subject changed from [tools] sporadic "Unable to clone Git repository" for wheels to [epic][tools] sporadic "Unable to clone Git repository" for wheels
- Description updated (diff)
- Assignee set to okurz
- Status changed from New to Blocked
- Copied to action #118903: Repositories for wheels should be cached added
- Status changed from Blocked to New
- Assignee deleted (
okurz)
- Target version changed from Ready to future
For now first continuing with some other subtasks in the parent #108527 before we can reconsider this
Also available in: Atom
PDF