Job start up times RunnerDeployment vs RunnerSet

[jespada]

Description:

Context:

I've been experimenting with both RunnerDeployment and RunnerSet for job execution. While both are functional, I've observed a noticeable difference in job start-up times between the two.

Issue:

I've noticed that jobs initiated through RunnerSet tend to start up slower compared to those initiated via RunnerDeployment when using webhooks.

Questions:

1. Is it expected for RunnerSet to have slower job start-up times when compared to RunnerDeployment? If yes, could you provide some insights into why this might be the case? (webhook vs long polling?)
2. Does it make sense to use webhooks with RunnerSet? If so, are there any best practices or configurations to optimize start-up times?
3. Is there a way to improve the start-up times for jobs using RunnerSet to be more in line with RunnerDeployment?

Additional Information:

• I'm running the latest release of both RunnerDeployment and RunnerSet
• They both run with the same specs.
• RunnerSet is deployed using runner-scale-set-controller chart and the runner-scale-set chart
• RunnerDeployment is deployed using the actions controller chart and a the runners with the webhook scaling method similar to what's described in the docs

Thank you for your time and assistance. Looking forward to some insights into this behavior!

[mahmud2011]

It's been a while you started the discussion and unfortunately no answer! Did you recently experiment this again?

[jespada]

Yes I can see it daily, I'm changing some of my config setup to see if the RunnerSet "issue" improves and will report back (it could be tied to the underlining VM startup times, but both of them run on the same node-pool..)