Workflow cancellation left pages deploy wedged

[sfreilich]

I recently had a problem where my pages deploy was stuck in a state where it would perpetually fail with:

Deployment request failed for 75152aed304ba9378fd527c226e9849ca39d5eda due to in progress deployment. Please cancel 4cb1cd27f0f1bceb7df2e4b7cb82785922dc978d first or wait for it to complete.

(Created a thread about that here: https://github.community/t/pages-deploy-wedged-incorrect-request-failed-due-to-in-progress-deployment/234793)

I eventually got the issue unwedged manually using the GitHub API to delete the relevant deployments. It seems what happened was a deploy job was cancelled midway through, but that cancellation failed: https://github.com/google/ink-stroke-modeler/runs/5457992551

Error: Error: Request failed with status code 404
{"message":"Not Found","documentation_url":"https://docs.github.com/rest"}
Sending telemetry for run id 1948901688
Deployment cancellation failed Error: Request failed with status code 401
[...]
url: 'https://api.github.com/repos/google/ink-stroke-modeler/pages/deployment/cancel/4cb1cd27f0f1bceb7df2e4b7cb82785922dc978d',
[...]

That URL seems to be coming from here:

deploy-pages/src/index.js

Line 20 in b405140

const pagesCancelDeployEndpoint = `https://api.github.com/repos/${process.env.GITHUB_REPOSITORY}/pages/deployment/cancel/${process.env.GITHUB_SHA}`

But I'm confused because I don't see that API endpoint in the documentation: https://docs.github.com/en/rest/reference/pages

And I wasn't able to get it to work myself with curl, I had to use the deployments endpoint to search for and delete the relevant deployments instead: https://docs.github.com/en/rest/reference/deployments#delete-a-deployment

Was this problem due to a bug in cancelHandler?

[DmitryGrayscale]

got the same today, but with error 400 (same description though: Deployment request failed due to in progress deployment)

[hsm]

Same issue here. I’ve created a minimal example here: https://github.com/hsm/pages/

Succeeds initially and then fails every time afterwards.

Steps to reproduce:

1. Create a public repository with an index.html of your choice.
2. Create a branch with the name gh-pages. This will automatically trigger deployment to Github Pages successfully.
3. Make a change in the index.html file on the gh-pages branch. This time the automatic deployments fails for me with an error message like this one:

Failed to create deployment for 48e85257726336084dffa6b610228d4ca7b1adbc.
{"message":"Deployment request failed for 48e85257726336084dffa6b610228d4ca7b1adbc due to in progress deployment. Please cancel f7ee6f59e1b777003098bf2f1c6f000ff04c4ec2 first or wait for it to complete.","documentation_url":"https://docs.github.com/rest/reference/repos#create-a-github-pages-deployment"}
Error: Error: Request failed with status code 400
Error: Error: Request failed with status code 400

[leemlb06pmi]

I'm getting this same problem today - potentially a process change we don't know about? Do we have any solutions as of yet?

[manuelmeurer]

Same problem here:
https://github.com/manuelmeurer/manuelmeurer.github.io/runs/5586680873

[thomiz]

Same issue in my public repo

[hassaanbinaslam]

Same issue here
https://github.com/hassaanbinaslam/myblog/runs/5586630206?check_suite_focus=true

[HavenTong]

Same issue here
https://github.com/HavenTong/HavenTong.github.io/runs/5586948035?check_suite_focus=true

[Ryo-N7]

same issue here:

https://github.com/Ryo-N7/Ryo-N7.github.io/runs/5587011624?check_suite_focus=true

[cnftstats]

Same issue here:

[eliasnemr]

same issue

[CesarDavidMorenR]

same issue but error 400

[AmaniZandalari]

same issue, error 400

[ThesimpleGithub]

same issue, error 400....

[FabianBartl]

same issue, but error 500

[kera-cudmore]

I'm currently having the same issue too, error 400

[rocketstack-matt]

I had this issue for the last couple of hours, but seems to have been related to this outage - https://www.githubstatus.com/incidents/sksd097hm0y5 . . . a blank change to force a redeploy of the gh-pages branch has worked having failed multiple attempts in the prior couple of hours.

[leemlb06pmi]

I got one deploy out the door - but the subsequent change failed with the same 400 error

[TienZhao]

Same issue since a few hours ago. Fail to deploy now. 😭
https://github.com/TienZhao/covid-shanghai/actions/runs/1999467222

[leo2105]

Same error here...
https://github.com/leo2105/mywebsite/actions/runs/1999598076

[kera-cudmore]

I got one deploy out the door - but the subsequent change failed with the same 400 error

Same here 😬

[mrbuche]

I think GitHub is having some issues today, and it's probably nothing we are ALL doing wrong...

[HavenTong]

I got one deploy out the door - but the subsequent change failed with the same 400 error

Same here.
https://github.com/HavenTong/HavenTong.github.io/actions/runs/1999597626 deployed successfully.
https://github.com/HavenTong/HavenTong.github.io/actions/runs/1999624535 the subsequent failed.

[ahmtcn123]

+1 Here

[RagingTiger]

Same: https://github.com/RagingTiger/tensorflowjs-mobilenet-transfer-learning/actions/workflows/pages/pages-build-deployment

[abubelinha]

Me too!
Thanks @sfreilich for opening this and linking back here.
@abubelinha

[counter-prop]

Same issue

[awagmi]

same here. i guess we just have to wait

[Clarit7]

Same issue. Like @mrbuche said, it should be GitHub problem.

[YiMysty]

👋 we have identified the potential issue due to a recent deployment since yesterday. We gonna deploy a fix soon. Currently pages deployment is 1 hour per repo per pages deployment. Will notified again when the fix released. Deeply sorry for the incident.

[Ivens-Zhang]

I just supposed that something wrong I had done to cause it.

[harrydehix]

Fixed for me now,

Just a matter of waiting?

I just tried rerunning the failed jobs again and again, and suddenly it worked. See here.

[Ivens-Zhang]

You are so lucky, I hope rerun failed job several times could work.

[TaylorLancaster12]

I'm also having this issue today on my public repo. Can't get my deploys to run, and all it is is CSS and HTML.

[izahn]

There are two community discussions about this:

• https://github.community/t/pages-deploy-wedged-incorrect-request-failed-due-to-in-progress-deployment/234793
• https://github.community/t/github-actions-failed-to-deploy-my-site-to-github-pages/239017/5

Looks like this issue is back as of 3/28/22 AM.

[vasveena]

I get the same issue. I deleted the existing deployment. After that, the deployment is stuck in "Current status: updating_pages".

[Lmmmmmm-bb]

I got same issue in this repo. always Current status: updating_pages.

[fffabiang]

I have the same issue. Stuck in Current status: updating_pages.

[rcanelav]

Same here :(

[GabrieLempert]

same here

[AlexandreAlvarez59]

Same here T.T

[aheze]

Anyone find a workaround? Same thing here

[Jean-Monsieur]

Same here :(

[ZoranPandovski]

Make sure you delete all your previous failing builds and just re-run the latest. Now, it should work

[vasveena]

Re-deployed without any changes. Works for me now!

[TaylorLancaster12]

Working now for me as well; deleted my failed workflows and deployed a new change with no issues.

[yoannchaudet]

Hello 👋,

Just to confirm, it does correspond to our incident this morning https://www.githubstatus.com/incidents/z95g34pl8rdm.

It took us longer than expected to get everything under control again and we identified a few things we will need to improve shortly so it does not happen again. Sorry for the inconvenience.

[Ivens-Zhang]

Hello 👋,

Just to confirm, it does correspond to our incident this morning https://www.githubstatus.com/incidents/z95g34pl8rdm.

It took us longer than expected to get everything under control again and we identified a few things we will need to improve shortly so it does not happen again. Sorry for the inconvenience.

It has already worked, Thanks for your contribution!

[LearningGp]

I met similar problems yesterday, after making some attempts that didn't work, I chose to put it on hold. When retrying this morning, the deployment was successful. There does seem to be an unfinished deployment task, but there is no way to cancel.

[orchetect]

This same issue is happening now on one of my repos. GitHub Status shows all green.

[yoannchaudet]

@orchetect We made changes in the last few days (both on the backend and on the client side to limit this scenario from happening. In the worse case a deployment should not block another one for more than 10 minutes (previously 60 minutes), in most cases it should not have to come to that.

[orchetect]

@yoannchaudet It took a few hours but did seem to resolve

[tex0l]

I'm experiencing the same issue right now on my repo: https://github.com/tex0l/tex0l.github.io/

[ToonTalk]

I intermittently began experiencing this on 7 December have using the Github Desktop for several years without any problems. I saw a suggestion to replace the Github Pages with None at github.com and then restore it to master. But now the entire github.io site has disappeared https://ecraft2learn.github.io/ai/

Can someone please tell me what to enter in the command line to restore gh-pages to the current state of the master repository?

UPDATE: When I reduced my repository from about 5GB to less than 2GB the problem went away. I wish the error messages were more informative.

[jeremy-daley-kr]

@yoannchaudet I saw you closed this. I'm working on a PR to convert the use of axios out to the @actions/github package, but don't see an answer to @sfreilich question, as I'm wondering it also...

Where is this endpoint?

const pagesCancelDeployEndpoint = `${this.githubApiUrl}/repos/${this.repositoryNwo}/pages/deployment/cancel/${this.buildVersion}`

I don't see it documented anywhere.

[yoannchaudet]

This API is still marked as internal. This is customary for features while they are in Beta so we can make breaking changes without having to go through a regular depreciation process.

Will make a note for the team to decide if we publish it now or hold a bit longer.

[Peter-developer01]

Same problem. Any ideas?

[fmaussion]

@Peter-developer01 I also landed here today which I think hints at server problems at github. For me waiting and re-triggering a build after wait seems to have solved the issue (15 minutes maybe?)