Hello,
I have deployed the application in GCP k8s, and I can view the objects created:
And I can view sometimes a new runner is added to my Bitbucket workspace:
The problem is that the runner is alwais in the "UNREGISTERED" state. I think the Controller and the Controller Cleaner are creating and destroying the runners, but the runners are not working properly.
I can also view the secrets created:
I have used the templates published here:
https://bitbucket.org/bitbucketpipelines/runners-autoscaler/src/master/config/
These are the Controller logs (most recent on top):
An these are the runner logs (most recent on top):
Finally, the pipeline never starts:
This is the pipeline definition:
___
___
Thank you very much for your help.
@Aurelio Pérez Noriega update.
I can confirm that the bug exists with runners always in UNREGISTERED state.
In some cases when green k8s jobs linked to non-existing runners (example when runners are manually deleted); and red k8s jobs linked to runners that UNREGISTERED and can't be in ONLINE state because k8s jobs are not ready.
As a workaround to solve this you could manually delete k8s green jobs, where non-existing runners linked (in logs they are always trying to change state to ONLINE with no success)
Regards, Igor
My purpose is to have an script to install all objects in k8s with no manual interaction, and to have the k8s jobs running OK in the Bitbucket Workspace. I think the workaround is not valid for my team.
Maybe you are close to have a solution to fix the problem. Can you give me an ETA to have a fixed version?
Thank you very much for your help.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
@Aurelio Pérez Noriega please contact our support team.
Regards, Igor.
Alternatively, i could suggest you to check:
- your k8s jobs are running (green status);
- in your k8s jobs there are logs presented with "waiting for runner ONLINE state"
- also try manually create runner via Runners page (without using this tool), and see if runner become in ONLINE state.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi @Aurelio Pérez Noriega .
So your runners in Bitbucket API never became in ONLINE state?
You should wait some minutes (1-2) after runner created to change state from UNREGISTERED to ONLINE.
Also provide us your runners-autoscaler-config configMap.
Keep in mind if you set to low value for
runner_cool_down_period
variable, cleaner will try to delete your runner before runner change the state to ONLINE.
Also for testing you can always start autoscaler without cleaner deployment, and create cleaner deployment later when your runners will be in ONLINE state.
Also the Controller logs are the most useful logs for us, please provide more information related to controller logs.
For testing you can lower value of
runner_api_polling_interval
to in example 60 (seconds) to check the status of your configuration more frequently.
Regards, Igor
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
The same logs in text format:
Thank you very much for your help.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.