equinix-2cpu-8gb does not scale

[stianst]

I've been experimenting with hosted runners using equinix-2cpu-8gb in cncf/keycloak-testing. However, it does not seem to scale the number of available runners.

Last week when I tried I hit a cap on 3 parallell runners, but this week it seems worse and I'm not getting more than one runner at a time.

I simulated there being 4 open PRs to Keycloak, and after about an hour I have 1 job running, with a queue of 145.

[stianst]

After 2 hours it's still only running with 1 job in parallell

[RobertKielty]

@stianst thanks for logging this issue.

@jeefy I will take a look at this

[RobertKielty]

@stianst please send me a refs to an action or workflow where you are testing this.

I need a link to both the action's YAML file and specific runs where there was an absence of scaling.

[stianst]

@RobertKielty https://github.com/cncf/keycloak-testing/actions?query=branch%3Aequinix

I started 4 workflow runs at the same time, and even with 4 runs I didn't see more than a single job running at any given time.

[RobertKielty]

Thank you for sharing this. Sorry, I missed this today, I will take a look first thing in the morning. Will DM you on Slack if I have questions.

[stianst]

Tried again today using a dummy workflow - it's basically just spanning 2 x 10 jobs, which sleeps for a configurable amount.

Running 3 runs with equinix-2cpu-8gb at the same time of this there's still a maximum of 1 job running at a time. Didn't wait for more than 5 minutes to see if any auto-scaling would kick in though.

@RobertKielty added you to the repo so you can try running the workflow yourself if that helps. Just open https://github.com/cncf/keycloak-testing/actions/workflows/dummy.yml and click on Run workflow, select the runner and time each job should wait for.

I've got a Java script that let's me pull how many jobs are running, but you can also just look at the status of the jobs through the GitHub UI.

[RobertKielty]

@stianst Thanks for adding me to the repo, I saw your update on the service desk request and I will discuss that with @jeefy and @krook tomorrow at the latest.

I will take a look at the test job and migrate it over to an Oracle Cloud runner as that is where we will be offering external runners moving forward.

[RobertKielty]

@stianst I modified the Dummy CI job to use a newly deployed hosted runner oracle-16cpu-64gb.

c. 8 jobs ran parallel, the elapsed time for entire run of 20 jobs was 7m 46s

https://github.com/cncf/keycloak-testing/actions/runs/13112604484

The runner runs on OCI, on a amd64 CPU (I will add amd64 to the runner name in time) For more runtime details, see our test-runner action here: https://github.com/cncf/automation/actions/runs/13059704463/job/36580556429

Can you use this runner to run full CI for Keycloak?

Let me know of any changes you would like to see on this runner.

cc @jeefy @krook