While setting up a Windows CI pipeline from GitLab, I was going through the numerous issues related to the Windows gitlab-runner docker executor that is using an old API (1.18) which Docker no longer accepts.
The issue results in the following error messages when the Gitlab/CI tries to connect to the runner:
Running with gitlab-runner 11.2.0 (35e8515d) on Windows VS2017 x64 0825d1d7
Using Docker executor with image buildtools2017 ...
ERROR: Preparation failed: Error response from daemon: **client version 1.18 is too old.** Minimum supported API version is 1.24, please upgrade your client to a newer version (executor_docker.go:1148:0s)
The 'buildtools2017' docker image that is referred to is the Microsoft "official".
The image seems to be working and valid for the current (experimental) Docker version I'm using (18.06.1-ce-win74) and for the stable version as well.
The issue was described throughout the GitLab wiki. Andrew Leech (?) went so far as to fork and modify the runner so that it would connect properly, and kindly provided his scripts and comments in a blogpost. This seems to give some results:
C:\gitlab-runner>gitlab-runner.exe -v
Version: 10.8.0~beta.551.g67a6ccc7
Git revision: 67a6ccc7
Git branch: windows-container-executor
GO version: go1.9.4
Built: 2018-07-30T08:57:44+00:00
OS/Arch: windows/amd64
The GitLab wiki states that they're waiting until a more stable solution can be released. Currently it's been over one year of broken windows docker runners..
Andrew's blogpost and a link to his gitlab-runner.exe describes actually a different workaround using the PowerShell runner that then starts a Docker instance. All the token info is exposed, I'm not sure how to set it up, and it also seems to rely on an external image with older build tools.
It seems the docker runner now connects, but if I undestand correctly, the Gitlab-runner docker runner does not seem to agree on the 'build directory' that is used. The first Gitlab/CI scriptline in my repo is just an echo command, so the error is not about the ci script content, but I'm not sure what it IS about. If anyone with docker fu knows what is going on this would really help me.
Using Docker executor with image buildtools2017 ...
ERROR: Preparation failed: build directory needs to be absolute and non-root path
Cheers,
Related
I am facing a problem with the Build error on circle-ci.
this is my repo
Error
Build-agent version 1.0.41563-0e4d6629 (2020-10-22T11:30:36+0000)
Docker Engine Version: 18.09.6
Kernel Version: Linux 2e2f534dcc94 4.15.0-1077-aws #81-Ubuntu SMP Wed Jun 24 16:48:15 UTC 2020 x86_64 Linux
Starting container circleci/node:6.14.8
Warning: No authentication provided, this pull may be subject to Docker Hub download rate limits.
image cache not found on this host, downloading circleci/node:6.14.8
Error response from daemon: manifest for circleci/node:6.14.8 not found
6.14.8 is a really historic version. Are you sure you didn't mean 14.6.8, 14.8.6, etc.?
In order to see what versions CircleCI supports, see pre-built CircleCI Docker images.
You can also check official node versions you can pull at https://hub.docker.com/_/node
Off-topic: you're not authenticating your pulls. Docker is applying pull rate limit as of Nov 1st, so it would be a good idea to authenticate before the pull, see Docker auth on CircleCI for a tip.
I have two runners: one on Linux works fine, but one on Windows has problems that I try to solve. In order to know the current state of the runner I use two commands: verify and status. When I run verify:
>gitlab-runner.exe verify
outputs:
Verifying runner... is alive ←[0;m runner←[0;m=c6xxxxxx
while status
>gitlab-runner.exe status
outputs:
gitlab-runner: Service is not running.
Question
What is a difference between being alive and running?
PS
This questions is not about why it is not running, it is about understanding the status.
gitlab-runner version
Version: 10.6.0
Git revision: a3543a27
Git branch: 10-6-stable
GO version: go1.9.4
Built: 2018-03-22T08:34:34+00:00
OS/Arch: windows/amd64
gitlab-runner. exe status tells you, if the Service is running. It seems that you've stopped it with gitlab-runner.exe stop.
gitlab-runner.exe verify tells you if your Runner ist still registered to a GitLab Instance. If you or someone else removed it in the GitLab CI/CD settings, your Runner would'nt be alive anymore.
I need to install GitLab on a server running Windows 7, but I'm blocked at this line. The documentation doesn't really helping me. The following is from my command prompt:
C:\GitLab-Runner>gitlab-runner.exe register
Please enter gitlab-ci coordinator URL (e.g. https://gitlab.com/):
https://gitlab.com
Please enter the gitlab-ci token for this runner:
Where can I find this token?
You're attempting to install the GitLab Runner which is used to run your jobs and send the results to a GitLab which is a server. As you're talking about GitLab running at a server you have to install that and not the Runner.
But it is not supported to install GitLab on Windows, see here in the GitLab forum. They recommend to use Linux in a virtual machine for that if you want it on Windows.
In all seriousness this is something that will probably never be supported.
Nevertheless to get the needed project registration token follow these steps described here. Also there is a discussion about it on GitHub.
To create a specific Runner without having admin rights to the GitLab instance, visit the project you want to make the Runner work for in GitLab:
Go to Settings ➔ Pipelines to obtain the token
Register the Runner
Further the process of registering the GitLab Runner which is actually what you're doing is described here.
As of last night all our new docker deployments started failing because the latest version of docker (docker-1.3.2-1.0.amzn1.x86_64) in the amazon repo fails to start up.
Steps to reproduce are:
## Launch instance with default amazon AMI
yum install docker-1.3.2-1.0.amzn1.x86_64
service docker restart
### Get the following error in /var/log/docker
2014/11/26 05:14:16 docker daemon: 1.3.2 c78088f/1.3.2; execdriver: native; graphdriver:
[8f6d7cfb] +job serveapi(unix:///var/run/docker.sock)
[info] Listening for HTTP on unix (/var/run/docker.sock)
docker: relocation error: docker: symbol dm_task_get_info_with_deferred_remove,
version Base not defined in file libdevmapper.so.1.02 with link time reference
If I downgrade back to docker-1.3.1-1.0.amzn1.x86_64 everything seems to be fine.
Is the AWS package actually broken, or is it just our setup?
Is there a work around other than downgrading?
Yes, it is broken for me too.
Downgrading has been the solution yet.
The same error was by me on a centos VM provisioned at my workplace - a yum update resolved it.
I suspect a build was broken but went out, and has been fixed subsequently.
I'm using Jenkins on Win7 and i've installed tomcat for ssh-agent plugin. And I could clone my GitLab project via git bash via ssh.
But if I build the project by Jenkins, it always says :
[ssh-agent] Using credentials IliptonChen(APRTest)
[ssh-agent] Looking for ssh-agent implementation...
[ssh-agent] FATAL: Could not find a suitable ssh-agent provider
FATAL:[ssh-agent] Unable to start agent
The full output text is here
Did I do anything wrong?
Check the version of your ssh-agent used by Jenkins.
This bug (for linux, but could apply to Windows too) reports (10 days ago, January 2014) this very same error message:
"JENKINS-20276: Native Library Error after upgrading ssh-agent from 1.3 to 1.4".
Downgrading to 1.3 resolves the issue.
Update 2019, five years later: as commented, this should be fixed now.
ssh-agent.exe is part of a Git for Windows distribution
D:\git\git>where ssh-agent.exe
D:\prgs\gits\current\usr\bin\ssh-agent.exe
(provided path/to/git/usr/bin is first in the %PATH% used by Jenkins)
Assuming you've installed Windows Git on Windows slave, it comes with ssh-agent binary (e.g. C:\Program Files\Git\usr\bin). Try adding its path to system variable PATH.
Otherwise untick SSH Agent and choose the credentials by selecting Credentials from dropdown in Source Code Management section.
Another way is to generate personal API token (OAuth) for that GitHub user and include that along with your repository address, e.g.
git clone https://4UTHT0KEN#github.com/foo/bar
For windows, the plugin still requires Tomcat to be installed in both master and slave.
I got this error because I was using an Ubuntu image for the agent, which doesn't have SSH installed.
agent {
docker { image 'ubuntu:focal' }
}
... so the solution was as simple as installing SSH as part of the pipeline:
steps {
sh "apt-get update && apt-get install ssh -y"
// rest of your steps here...
}
In my case, the error was accompanied by an error about disk space depletion:
[ssh-agent] FATAL: Could not find a suitable ssh-agent provider
[ssh-agent] Diagnostic report
[ssh-agent] * Exec ssh-agent (binary ssh-agent on a remote machine)
[ssh-agent] hudson.AbortException: Failed to run ssh-agent: mkdtemp: private socket dir: No space left on device
So I ran docker system prune -a, which fixed it.