unfortunately i need to run a pipeline inside my cluster but right now it's not possible to due SSL problem that basically it seems to me very strange @Justin Thomas
This great news, as it's a feature I've greatly missed from GitLab.
However, I can't figure out from the existing docs how to specify a particular image to be used in my pipeline on my runner. For example, I tried to run a SonarQube scan using the SonarQube container and using the `runs_on` configuration in my pipeline:
The existing docs only provide an example using a custom: pipeline, and the docs for that seem to imply that it is only for use with manually triggered steps, which is not at all what I want. The above configuration seems logical and intuitive, but it's not working, and I can't figure out why it's not working. The docs provide no insight on how to troubleshoot or debug Bitbucket <-> Runner communications. The runner shows up as "online". But there seems to be no indication that it's being sent a job at all. The SonarQube container is definitely being used, but it seems as if it's being launched on Bitbucket's runners, not mine.
Is there more complete documentation available somewhere?
Atlassian Team members are employees working across the company in a wide variety of roles.
May 26, 2021 edited
@Luis Thanks for trying out the runner. As self-hosted runners is in beta the REST APIs are still internal, we will be making it public when we go GA. I would suggest not depending on the internal APIs till then. Thanks
Atlassian Team members are employees working across the company in a wide variety of roles.
May 26, 2021 edited
@polRk There are multiple things that can impact the speed of a step on a self-hosted runner. Network latency is a known cause to slow down builds in self-hosted runners. I would suggest getting help from Atlassian support to debug the issue. Thanks for trying the runner.
What do you mean macos runner? We dont currently support macos are you meaning running it under docker for mac?
The first time you run a step it will have to pull the images from the docker registries which may not be geo located close to you, same for artifacts and caches.
In the cloud we have a registry mirror which caches most of the public layers of dockerhub, and we are geo located close to dockerhub and where caches and artifacts are stored plus the networking provided to those instances would be much higher than whats available to your development machine.
If you would like us to investigate please raise a support ticket, so we can advise.
Your yml file is incorrect you need to use `runs-on` not `runs_on` as we ignore unknown properties its thus not reading your runs_on and scheduling on our cloud runner which is the default behaviour when a step doesnt contain `runs-on`.
Will users be able to allocate more memory to the `docker` service in a self-hosted runner?
Builds are currently constrained by the same 1x and 2x limits that exist with the Atlassian infrastructure, and my motivation for using a self-hosted runner is to allocate more resources to builds.
Is anyone else having issues seeing logs from the execution on the local runner appear in the Bitbucket UI? I see the 'Runner' and 'Build teardown' sections, but nothing in between. If I comment out my 'runs-on' line and have the pipeline execute in the Bitbucket cloud then the log entries show up as expected.
Update 2021-06-10: have raised a ticket and the issue has been referred to the internal development team after some troubleshooting efforts.
Just trying to get a runner working on a RHEL7 docker host and getting java errors related to a missing lib
[2021-06-11 09:25:37,385] Runner starting version: 1.139 [2021-06-11 09:25:38,537] Exception encountered during context initialization - cancelling refresh attempt: org.springframework.beans.factory.UnsatisfiedDependencyException: Error creating bean with name 'dockerServiceImpl' defined in URL [jar:file:/opt/atlassian/pipelines/runner/runner.jar!/com/atlassian/pipelines/runner/core/service/docker/DockerServiceImpl.class]: Unsatisfied dependency expressed through constructor parameter 0; nested exception is org.springframework.beans.factory.UnsatisfiedDependencyException: Error creating bean with name 'containerServiceImpl' defined in URL
Then buried down in the hundreds of lines of Java stack trace I can see:
Caused by: java.lang.UnsatisfiedLinkError: could not load a native library: netty_transport_native_epoll_x86_64 at io.netty.util.internal.NativeLibraryLoader.load(NativeLibraryLoader.java:226) at io.netty.channel.epoll.Native.loadNativeLibrary(Native.java:250) at io.netty.channel.epoll.Native.<clinit>(Native.java:69) at io.netty.channel.epoll.Epoll.<clinit>(Epoll.java:39) ... 68 more
Anyone else seen similar errors and know what the required RHEL rpm is to supply netty_transport_native_epoll_x86_64
I thought I had found it with devtoolset-4-netty.noarch but this seemed unlikely as a noarch rpm is unlikely to deliver a x86_64 and didn't resolve the issue
We are currently looking at switching to git from SubVersion and I have been trialling GitHub, GitLab and BitBucket, have got self-hosted runners working on the other 2 and want to get it working on BitBucket as well
@Thomas Varga Have set up a CentOS 7 box with podman installed, what did you put in /etc/containers/registries.conf to configure the Atlassian registry
The docker command to run had docker-public.packages.atlassian.com/sox/atlassian/bitbucket-pipelines-runner as the name of the image
I tried just adding docker-public.packages.atlassian.com as follows
@Alec Keeler, I may have inadvertently caused some confusion. On CentOS 7 I am using Docker. RHEL however uses podman to run containers. The former I have working with a BB runner , the latter I have not tried.
To get the runner to work on CentOS/Docker I followed the steps in https://support.atlassian.com/bitbucket-cloud/docs/set-up-and-use-runners-for-linux/ . When adding a runner in the BB UI you will be provided a CLI command to copy and paste. Use that verbatim. That will download the Atlassian image and also register your container with the BB cloud. As long as you have the Docker service running it should work without any extra Docker config changes. I followed the instructions on the page to the letter. Apart from an issue with logging all else appears to be working.
@Justin ThomasPreviously you have shared that restrictions of memory/run-time will be removed on Pipeline Runners. However, what will happen with the costs? Will you still charge per build minute when using Pipeline Runners? I would expect costs to be significantly lower as we are using our own hardware for this and. Any info about this?
"Runners allows you to run builds in Pipelines on your own infrastructure, and you won’t be charged for the build minutes used by your self-hosted runners."
@Justin Thomas there is something strange, I can run the build only once, all subsequent builds are hanging. To recover I have to stop, rm and run the docker again.
Changed to use docker from dockers own repo on CentOS 7 (installed as per https://docs.docker.com/engine/install/centos/) and the docker runner image now starts OK and connects to bitbucket
Think there might be an issue with the RedHat packaged version of docker included in RHEL 7, so be aware if using the latest RedHat rpm, docker-1.13.1-208.git7d71120.el7_9.x86_64.rpm
Just an FYI I was able to get this working on Ubuntu using SNAP by altering the docker command as the location of /var/lib/docker is moved when running it under SNAP.
I would like to put a vote to have both repo specific runners as well as runners that were available for all repos in our Workspace. Our use case is that we want to access an on-prem Thycotic Secret Server to be able to pull passwords dynamically during our deployment. To have to run a container for every repo would make this unwieldly.
Love the new feature, is there a public road map available?
113 comments