Open mikhailkoliada opened 1 month ago
@MikeMcC399 yep, I forgot about it, now done :)
Run java -version java -version chmod +x ./gradlew ./gradlew build fatJar --no-daemon java -jar build/libs/headlong-cli-1.1-SNAPSHOT.jar -version shell: /usr/bin/bash -e {0} env: GRADLE_ACTION_ID: gradle/actions/wrapper-validation JAVA_HOME: /opt/hostedtoolcache/Java_Zulu_jdk/8.0.412-8/x86 JAVA_HOME_8_X86: /opt/hostedtoolcache/Java_Zulu_jdk/8.0.412-8/x86 The futex facility returned an unexpected error code. /home/runner/work/_temp/f9c9c83d-d717-4c5b-9083-6922b5f9175c.sh: line 1: 2720 Aborted (core dumped) java -version Error: Process completed with exit code 134.
convert
(imagemagick) is no longer pre-installed.
See run https://github.com/agda/agda/actions/runs/9086817771/job/24973341288?pr=7261#step:7:105
WARNING: Unable to run the image conversion command 'convert'. 'sphinx.ext.imgconverter' requires ImageMagick by default. Ensure it is installed, or set the 'image_converter' option to a custom conversion command. Traceback: [Errno 2] No such file or directory: 'convert'
@mikhailkoliada Thank you for the release! What is the timeline for updating the ubuntu-latest
alias to default to Ubuntu 24.04 instead? (Context: We would like to upgrade to the latest Ubuntu version, however, it will save a lot of churn across many repos of replacing runs-on: ubuntu-latest
with runs-on: ubuntu-24.04
if we could instead wait for the default to change.)
@edmorley
In https://github.com/actions/runner-images/issues/9691#issuecomment-2063926929 there was mention of a date in August 2024 for GA. That might coincide with the change you are asking about.
Are the changes marked "Removed from the Ubuntu 24.04 image due to maintenance reasons." final? Is there a particular common reason? Can we somehow vote/plead to revert any of those choices?
I expect a lot of builds with "ubuntu-latest" to break if it hits GA in that form.
@kojoru there will definitely be changes until GA including the software list installation (for instance we are to add firefox & edge, but a bit later), but we do not guarantee we are to address all the requests, everybody is welcome to file new tools addition requests and we'll review them.
@edmorley migrating to lates is a different process that usually takes time, 24.04 will eventually become latest after its GA
@mikhailkoliada Would you like to amend the software difference table in your first post with the information that Podman gets updated from Podman 3.4.4 in the Ubuntu 22.04 image to Podman 4.9.3 in 24.04?
@mikhailkoliada Cannot seem to use the 24.04 image:
@sanderdelfos
I noticed also that ubuntu-24.04
does not yet appear on the Azure Pipelines > Microsoft-hosted agents list.
@MikeMcC399 True, but it is listed in this issue as ready for use and also listed here: https://github.com/actions/runner-images
I noticed Ubuntu2404
was not added as an option to helpers/GenerateResourcesAndImage.ps1
. Are you holding off until GA or are there any plans to do that earlier?
Hello,
I want to use ubuntu-24.04 in my release pipelines but I don't see this in the "Agent Specification" drop down. How can I get it ? What is my alternative ? Currently I am able to use 24.04 in my azure-pielines.yml used in build pipelines to build image.
Thank you in advance.
@pmandadkes1207 It seems that Ubuntu 24.04 is not fully deployed on Azure DevOps, @sanderdelfos and I can't even use it on a YAML pipeline (but you can). Let's give it a bit more time :)
Ruby
and Python
). The most complete state of the image can be found here. And please keep in mind that the situation will change quite quickly, since the availability of packages at the start of a new image changes most quickly.@f-lawe Sorry for the oversight on my side. I just realized that I was using 'ubuntu-latest' in azure-pipelines.yml which points to 22.04. I am gettig the same error as @sanderdelfos when I explicitly mention 24.04.
Over the day I started to run into an issue in azure devops regarding "azure cli" after this ubuntu latest build, with the task AzureResourceManagerTemplateDeployment@3
Starting: AzureResourceManagerTemplateDeployment
==============================================================================
Task : ARM template deployment
Description : Deploy an Azure Resource Manager (ARM) template to all the deployment scopes
Version : 3.238.0
Author : Microsoft Corporation
Help : https://docs.microsoft.com/azure/devops/pipelines/tasks/deploy/azure-resource-group-deployment
==============================================================================
ARM Service Connection deployment scope - Subscription
Checking if the following resource group exists: **REDACTED**.
Resource group exists: true.
Creating deployment parameters.
##[error]Check out the troubleshooting guide to see if your issue is addressed: https://docs.microsoft.com/en-us/azure/devops/pipelines/tasks/deploy/azure-resource-group-deployment?view=azure-devops#troubleshooting
##[error]Error: Command failed: az version
/bin/sh: 1: az: not found
Finishing: AzureResourceManagerTemplateDeployment
Is this an expected thing to be after the release? Or an unexpected breaking change?
Hey @iXyles!
ubuntu-24.04
<> ubuntu-latest
Your issue is related to ubuntu22.04
.
@erik-bershel Thanks for the notice! I was mixing the numbers up...
Is the additional comment in the PR above enough to get a notice on this, or should I raise a new ticket?
@iXyles you may file a compatibility issue here: https://github.com/microsoft/azure-pipelines-tasks - it's a task owners repository. We won't accept this bug since base functionality of azure cli
available. See there:
Massive thanks @erik-bershel for taking the time to verify this, will look into doing that then!
EDIT: Sorry for misstag.. wrote on phone
Would it be possible to have Python 3.8 as a cached version? I could understand not having 3.7, since it's end of life, but 3.8 is maintained until October 2024, so having it in the cache as well could be nice.
wine seems to crash with futex error
The futex facility returned an unexpected error code.
Aborted (core dumped) wine ./mpv.com -v --no-config
It happens also multiple times during apt-get, dpkg configuration, stage.
The futex facility returned an unexpected error code.
Aborted (core dumped)
Hey @erik-bershel. I'm one of the Cabal maintainers (the Haskell build tool included in the runners). The upgrade to GHC 9.10 was premature: nothing in the ecosystem is ready for it yet. Please, roll back to GHC 9.8 or better yet 9.6, which was declared to be a sort of an LTS by the GHC team recently.
Context. Currently, 9.10 is considered a bleeding edge. People can easily get it via ghcup (also included in the runners) but it's a bad default because, as @simonmichael mentions above: most software will fail to build with it. Cabal's own CI had a failure because we relied on the GHC from the runner for our "quick" jobs: they're supposed to run quickly, so we did't want to install any particular GHC and thought the the runner will provide a sensible default (https://github.com/haskell/cabal/pull/10026).
To add to the previous comment: GHC 9.10 works best with cabal 3.12 which has not been released yet.
Hey @ulysses4ever and @andreasabel!
Thank you for bringing it into our attention. Lets discuss it. Currently we have next process based on the previous conversation with @andreasabel :
available_versions=$(ghcup list -t ghc -r | grep -v "prerelease" | awk '{print $2}')
# Install 2 latest Haskell Major.Minor versions
major_minor_versions=$(echo "$available_versions" | cut -d"." -f 1,2 | uniq | tail -n2)
for major_minor_version in $major_minor_versions; do
full_version=$(echo "$available_versions" | grep "$major_minor_version." | tail -n1)
echo "install ghc version $full_version..."
ghcup install ghc $full_version
ghcup set ghc $full_version
done
As I see 9.10
is declared as a latest
non-prerelease
version in ghcup list -t ghc -r
. I think we may exclude latest
if you claim that the community will only benefit from it. It's also possible to filter versions to install in a way to get recommended
and latest base non-latest
non-prerelease
versions. Again, if it suits best the community needs. I have some vague memories that such a scheme has possible side effects due to the fact that the recommended
and latest
versions may be incompatible, but these may be false memories.
cc @mikhailkoliada
Listen to the GHC/Cabal maintainers over me, but :
Using what ghcup deems recommended would sound like the most reasonable thing to do, except ghcup has, in my opinion, a very unintuitive and too conservative strategy for updating those. For GHC, ghcup-recommended makes some sense. For cabal, ghcup-recommended version was way out of date for a long time. So, my hunch would be to use ghcup-recommended GHC and ghcup-latest cabal (note, that generally using newer cabals with older GHCs is perfectly fine, but not vice versa). I'm fairly relaxed about what particular GHC version to get (9.4//6/8 but certainly not 9.10), though, but creating a general policy is hard...
@erik-bershel wrote:
available_versions=$(ghcup list -t ghc -r | grep -v "prerelease" | awk '{print $2}') # Install 2 latest Haskell Major.Minor versions major_minor_versions=$(echo "$available_versions" | cut -d"." -f 1,2 | uniq | tail -n2)
I think it would make sense to preinstall latest..recommended
(or, if they coincide, the latest 2), and then set recommended
as the default ghc
. In the current situation, ghc-9.10.1
(latest), ghc-9.8.2
, ghc-9.6.5
and ghc-9.4.8
(recommended) would be available and ghc
would point to the last of these.
Oh, if pre-installing several versions is an option, then what Andreas suggested looks perfect to me.
For our Ubuntu 24.04 runners, we are getting the warning Received request to deprovision: The request was cancelled by the remote provider.
followed by a cancellation of the run when it passes more than 4 minutes. Here is an example run.
I believe it started yesterday/today, and nothing noteworthy has really changed in our repository in the past couple of days.
Perhaps related to #7897.
I'm also getting Received request to deprovision: The request was cancelled by the remote provider
on Ubuntu 24.04 jobs after 4:10 seconds.
Sometimes it's failing at 4:10 but without the deprovision message and just The hosted runner encountered an error while running your job. (Error Type: Failure).
Example without deprovision message
I think there might be something resourcey going on as I've managed to get the job to run a couple of times when giving it less to do or running with less threads.
I am seeing another issue from azure ubuntu-latest build agent when I am building my python code from python 3.11.9 source using ubuntu 24.04 as base image.
22.30 ERROR: THESE PACKAGES DO NOT MATCH THE HASHES FROM THE REQUIREMENTS FILE. If you have updated the package versions, please update the hashes. Otherwise, examine the package contents carefully; someone may have tampered with them. 22.30 unknown package: 22.30 Expected sha256 xxxxx 22.30 Got yyyyy
I am not posting actual values but in my case during build. I am making python 3.11.9 as alternate install as ubuntu 24.04 by default supports only 3.12.
I wasn't seeing this issue with 20.04 which were my previous ubuntu and azure build agent versions. This is happening frequently but it's not consistent. Re-triggering the build will work.
Any idea what may be going on here ?
Regarding
Received request to deprovision: The request was cancelled by the remote provider
In our case, this happens when running multiple builds for multiple architectures in parallel, where APT packages are installed and upgraded within a QEMU-emulated container. To me it looks like the above warning appears for those jobs, which are currently printing/doing the
Calculating upgrade...
of APT, while installing build dependencies and upgrading the container image within the same call (apt-get dist-upgrade package1 package2 ...
). All other currently running jobs of the same workflow run are then cancelled as well, no matter what they are currently doing, producing the
The hosted runner encountered an error while running your job. (Error Type: Failure).
error instead.
Surely APT calculating upgrades is CPU-intense already, probably additionally with the new APT resolver, and multiplied when being QEMU-emulated. Quite an argument for the native ARM runners, which will reduce load over hosted GitHub runners massively, once users adopt. We will ASAP 🙂.
However, the same stuff worked and works well on Ubuntu 22.04 runners, and of course a proper reaction would be to limit CPU utilisation of the runner, instead of just killing it. RAM usage btw, as indicated in one of the older linked issues, is likely not the reason, since this produces different runner-side error messages, usually just crashing whichever RAM-intense step/command currently runs and/or OOM kills, but all transparent in logs.
This, unless the VM host itself runs out of RAM. Proper reaction then would be to reduce the runners assigned RAM, or the amount of available runners. Or upgrade host hardware, of course 🙂.
EDIT: Now faced the same when checking a larger number of shell scripts with shellcheck
, without any emulation.
I am seeing the same problem as @MichaIng but it only started showing up in the last day or so. Reverting to 22.04 for now until this is fixed.
UPDATE: This now seems to be affecting ubuntu-22.04 runners too. Wondering if there is a larger issue with GitHub hosted runners.
For our Ubuntu 24.04 runners, we are getting the warning
Received request to deprovision: The request was cancelled by the remote provider.
followed by a cancellation of the run when it passes more than 4 minutes. Here is an example run.I believe it started yesterday/today, and nothing noteworthy has really changed in our repository in the past couple of days.
Perhaps related to #7897.
Our runners now seems to be okay.
I am a maintainer of ORAS project. Will ORAS CLI still be reserved in Ubuntu 24.04 GA runner machine?
@mikhailkoliada, Swift just added 24.04 support for the latest patch release, would be good to add that to this image. Unless you meant something else by "maintenance reasons" above, other than there not being official 24.04 builds till now?
![Uploading fotor_1718323087600.png…]()
@mikhailkoliada
It would be great to see Firefox and Edge browsers added to the Ubuntu 24.04 beta > Browsers and Drivers section.
Is there any timeline available for this addition? I'm assuming that they are planned for the GA release of the runner in any case. Hopefully that is true?!
Will this image be available on the ARM64 runners?
Edit: thanks! https://github.blog/changelog/2024-06-24-github-actions-ubuntu-24-04-image-now-available-for-arm64-runners
thx
I've found that Ubuntu 24.04 image is not working for Azure DevOps self-hosted VMSS: https://github.com/actions/runner-images/issues/10122
gradle still failing with exit code 134:
https://github.com/esaulpaugh/headlong-cli/actions/runs/9684527870/job/26722612613
Tried upgrading but Firefox is missing. E.g. when using QUnit and Testem.
Launcher headless firefox not found. Not installed?
https://github.com/Krinkle/node-colorfactory/actions/runs/9734886433/job/26863453598
Working around by switching back to the Ubuntu 20 image for now.
Thanks
Breaking changes
Ubuntu 24.04 is now available for all GitHub Actions and Azure DevOps users. You can use the ubuntu 24.04 image label in your YAML to select this image.
GitHub Actions
Azure DevOps
The image is marked as "beta" for now. It means some software can be unstable on the new platform. Also, there could be queueing issues as the capacity will be balanced only throughout the next weeks. Please report any problems with the new image to this repository.
Platforms affected
Runner images affected
Software difference between Ubuntu 22 and Ubuntu 24