Open lucaslosi opened 1 year ago
Please check all (including closed) issues before you post.
Please check all (including closed) issues before you post.
kubernetes/kubectl#87
This isn't helping. We've tried 1.28.3 which is https://storage.googleapis.com/kubernetes-release/release/stable.txt
Today We've tried running 1.28.2 and 1.27.6 of kubectl with the same errors.
I am not in the same organization as him, additionally the issue is sporadic which suggests configuration drift on the runners.
use with: version: "v1.28.2"
works consistently, like they said use a "v" doh!
FWIW: I was able to get it to work by adding a v to my version, like so 'vX.XX.XX' in my workflow:
It failed here:
- uses: azure/setup-kubectl@v3
id: install
with:
version: '1.28.2'
And was successful here:
- uses: azure/setup-kubectl@v3
id: install
with:
version: 'v1.28.2'
Please check all (including closed) issues before you post. kubernetes/kubectl#87
This isn't helping. We've tried 1.28.3 which is https://storage.googleapis.com/kubernetes-release/release/stable.txt
Today We've tried running 1.28.2 and 1.27.6 of kubectl with the same errors.
I am not in the same organization as him, additionally the issue is sporadic which suggests configuration drift on the runners.
use with: version: "v1.28.2"
works consistently, like they said use a "v" doh!
How is my reply not helping.
1) you are opening the same issue that was closed today 2) My example literally mentions a "v" in the version number
It's very inconsistent today. had 5 workflows working and like 20 with errors.
I ran a curl command on the same url inside the workflow and it downloaded fine, so I think the issue can be related to the github toolcache: https://github.com/Azure/setup-kubectl/blob/dd44a92c26432fb9f47ec9c4a94fa40059fe2e32/src/run.ts#L57
setting the version with the 'v' worked for me, but i think the action should work with the default configs
I have a workflow that does not specify a version. It has been working fine for months. Now it fails with the error message shown in the original post here.
got the same issue - let us know when there is a fix. not specifying a version. default was working fine up until a few hours ago. would be great if you continue to support default (no version specified). I ran with "v1.28.3" and got it to work, but no version specified was working until earlier today.
Just as an FYI, I opened an issue on the kubectl issue tracker (https://github.com/kubernetes/kubernetes/issues/121400). Maybe some release update failed and now the stable.txt contains a version that doesn't correspond to a binary.
That's why default doesn't work, although it should probably use the fallback version specified in the script, and it just fails rather than using the fallback which would alleviate some pain points since all of my projects don't specify a pinned version.
I could make the PR that fixes the fallback version, but not sure if the authors see that as an issue.
We have pinned the version to v1.28.3
, but it still doesn't work for us:
Looking at the final lines:
##[debug]Downloading https://dl.k8s.io/release/v1.28.3/bin/linux/amd64/kubectl
##[debug]Destination /home/runner/work/_temp/81868074-d646-4f0e-932e-7814c84d2d82
##[debug]Failed to download from "https://dl.k8s.io/release/v1.28.3/bin/linux/amd64/kubectl". Code(404) Message(Not Found)
Error: Error: Kubectl 'v1.28.3' for 'amd64' arch not found.
I click on the URL and succeed in downloading the binary. Using curl -Li
, I see that I get HTTP/2 302
and then HTTP/2 200
. No 404 here.
At this point, it really feels like it's a GitHub Actions cache thing, even though I can't find where such thing is configured (I'm not very familiar with the actions/toolkit repo).
Anyone still having this issue after pinning the version and has additional thoughts or suggestions?
Facing this issue.
Action called without arguments:
jobs:
my-job:
runs-on: ubuntu-latest
steps:
- uses: azure/setup-kubectl@v3
Log shows:
Run azure/setup-kubectl@v3
with:
version: latest
Error: Error: Kubectl 'v1.28.3' for 'amd64' arch not found.
Edit; resolved for us after specifying a pin that isn't v1.28.3
, as suggested/alluded to earlier in this thread.
Facing this issue.
Action called without arguments:
jobs: my-job: runs-on: ubuntu-latest steps: - uses: azure/setup-kubectl@v3
Log shows:
Run azure/setup-kubectl@v3 with: version: latest Error: Error: Kubectl 'v1.28.3' for 'amd64' arch not found.
It makes sense, latest
is the default, and v1.28.3
is indeed the latest version. Problem is with downloading the binary.
Anyone still having this issue after pinning the version and has additional thoughts or suggestions?
We pinned to v1.28.2 instead, and everything seems to work. This evidence supports my assumption that it's an issue with cache while downloading the binary.
Anyone still having this issue after pinning the version and has additional thoughts or suggestions?
We pinned to v1.28.2 instead, and everything seems to work. This evidence supports my assumption that it's an issue with cache while downloading the binary.
Yeah, so what most likely happened here is that they released v1.28.3, updated stable.txt but the binary failed to build. So when they rectified the situation they uploaded the binary to GCS but some third parties (actions for example) still cache the 404.
last night the pinned approach stopped working in our CI. Not sure why yet.
i faced with this issue today. i rerun job while read this topic. first and second is stopped with subj error, but third is successfully finished.
i don't have any idea =\
This is still occasionally failing - some builds succeed, some fail. Bad cache somewhere?
I have a slew of failed jobs today.
This part constructs the URL for kubectl: https://github.com/Azure/setup-kubectl/blob/main/src/helpers.ts#L15
As of 2023-10-19, the default version translates to https://dl.k8s.io/release/v1.28.3/bin/linux/amd64/kubectl
. On one computer, I get a 302 followed by a 200. On another computer, I get a 404. Therefore, I think it's likely to be a CDN cache issue on dl.k8s.io, and not a problem of this setup-kubectl
module.
This part constructs the URL for kubectl: https://github.com/Azure/setup-kubectl/blob/main/src/helpers.ts#L15
As of 2023-10-19, the default version translates to
https://dl.k8s.io/release/v1.28.3/bin/linux/amd64/kubectl
. On one computer, I get a 302 followed by a 200. On another computer, I get a 404. Therefore, I think it's likely to be a CDN cache issue on dl.k8s.io, and not a problem of thissetup-kubectl
module.
You can reach the same generated URL running curl inside the same job that fails the setup-kubectl, so it's probably related to a manual file upload as @thecodeassassin said or the actions/toolkit cache
Makes sense. It should start to clear up soon
This issue is idle because it has been open for 14 days with no activity.
Same problem occurs when you use the newest v4.0.0. Downgrade to version 3 worked for me:
- name: install kubectl
uses: azure/setup-kubectl@v3
What happened?
I'm running the following step on my github workflow and got an error telling that the version '1.28.2' for 'amd64' was not found.
i've tried to downgrade the version adding the following, but got the same error:
The error is strange because reaching the download link locally on my machine works correctly.
Version
Runner
ubuntu-latest
Relevant log output