-
Notifications
You must be signed in to change notification settings - Fork 199
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Use incluster config fallback for KubernetesProvider #3357
Use incluster config fallback for KubernetesProvider #3357
Conversation
@benclifford do you have any thoughts on this CI failure? It also occurred on your PR #3359.
|
some exciting new build problem with radical's build/packaging. I tried briefly to recreate it on my laptop and did not have a problem, so I'll dig into it more deeply later in the week - I'm not working on Parsl today. |
(very possibly related to build system changes in Python 3.12 which were/are fairly aggressive) |
Oof. Thanks so much for providing that context and saving us from spinning our wheels on this! ❤️ |
@trey-stafford I don't have rights to update your branch, so I can't merge the latest parsl master (and in general that will make it hard to get your branch up to date to merge it later). But, recent PR #3360 should fix the radical test failure you reported. |
82c02a1
to
9f63b39
Compare
@benclifford I've merged in master. Happy to do that again as needed! Note that I also amended my first commit to this PR to properly credit @manning-ncsa for work contributed in the superseded PR, #2503 . |
@trey-stafford if you edit this PR can you enable "Allow maintainers to push"? Or is that only settable when opening the original PR? 🤔 |
you did something weird perhaps with rebase that makes the diff look weird - I think I know what, and I think that will fix itself once master gets merged again - which is necessary for putting this PR on master anyway as our tests require the PR be up to date. |
I think I see what happened, we'll get it sorted on Monday! Thanks for your patience :) |
9f63b39
to
0e1fc95
Compare
🚀 Sorry didn't get this done on Monday, but I think we're all fixed up now :) |
Can you comment a bit on what the exception reported to a user looks like? Say the user intends to do something with the config loaded by load_kube_config() but they mess it up in some way that raises an exception... does useful debugging information then get lost because that exception is immediately discarded? |
I have a couple meetings coming up to prepare for but we will get back to you in a couple hours! |
Would you like that context in the commit message for this commit? 4833fc2 |
It doesn't really matter where you write it - if you have something really interesting to say, a comment in the code would be ok too, for example. Just more interested that someone thinks about it because I am always suspicious of too-easily-discarded exceptions. |
549987a
to
293fba0
Compare
We pushed another change that we feel will help prevent a really confusing user experience if both local config and incluster config fail to load. We used an ExceptionGroup to handle this, but perhaps the better solution is to only re-raise the local config exception with a note that we tried both loading local and incluster config. The ExceptionGroup looks like this when raised:
And as we're typing this @trey-stafford called out that
|
Co-Authored-By: Trey Stafford <[email protected]>
Co-Authored-By: Trey Stafford <[email protected]>
fbbd8d9
to
67cdca4
Compare
Co-Authored-By: Trey Stafford <[email protected]>
@benclifford we think we've got things ready to go! |
ok cool. you still need to do something to let me update this branch, to fit in our merge process |
wrt your comment on using Python 3.11 features, in recent years our informal policy has been to support the same Python versions as are listed here as "bugfix" or "security", so under that model, no 3.11 features until near the end of 2026 - https://devguide.python.org/versions/ We have a low-key project going on to understand our usage a bit better, and my gut feeling is that with people using things like |
Shoot. I think @trey-stafford has to do that, and he may not be available until Monday. I don't have an edit button even though the PR is from a repo I have write access to. As ugly as it is, I can create another fork and open a new PR and ensure the "allow maintainers to change" checkbox is ticket if you'd like to get this merged sooner than that. Trey, when you get to it, this checkbox is towards the bottom of the right sidebar: |
@mfisher87 no super rush to mess around to get it merged |
This option is not available. I think it's because this is an org-forked repo. This documentation suggests it is possible with "user-owned forks". Maybe a new fork is the right path? |
Ok, I transferred the repo to my personal account and was able to enable edits and access by maintainers! @benclifford hopefully that does the trick! |
In retrospect, that makes sense, as checking that box would give maintainers of other projects access to org-wide secrets! 🤯 (there is no such thing as personal-account-wide secrets) |
ok that worked, so this will be in the next Parsl release happening in the next few hours. Thanks! |
👋 Hi @trey-stafford and @mfisher87! My name is Sophie and I'm Parsl's community manager. We're trying to broadly understand how people are using Parsl and would appreciate it if either/both of you would fill out our quick survey: https://bit.ly/parsl-use-survey. Thanks in advance! |
Hey @sophie-bui thanks for asking for our input! I just filled it out. |
Description
This PR supersedes #2503
See kubernetes-client/python#1005 for a good description of the issue.
Changed Behaviour
The
KubernetesProvider
now falls back to loading config in-cluster if a kube-config file is not found. This allows in-cluster submission of parsl jobs.Fixes
Fixes # (issue)
Type of change
Choose which options apply, and delete the ones which do not apply.