Not getting expected behaviour with pool config

[ghimire]

delay_webhook_event = 0
enable_ephemeral_runners = true

runners_maximum_count = 5

# override scaling down
scale_down_schedule_expression = "cron(* * * * ? *)"

# pool usage
pool_runner_owner = "xxxx"
pool_config = [{
  size                = 1
  schedule_expression = "cron(* * * * ? *)"
}]

enable_job_queued_check = true

minimum_running_time_in_minutes = 1

Hi, with the above config i'd expect to have at maximum 5 runners (4 event triggered runners + 1 pool runner). I'm hoping that at any moment there is 1 runner in the pool waiting for job. Furthermore, I'm expecting that if there are more jobs that require more than 1 runner in the pool, they get created but never exceeding 5. In an idle state where there are no events, there should only be 1 pool runner.

The reality is different however. Every minute the pool lambda keeps adding a runner to the pool and the number of runners just increase by minute. Is something wrong with this config? How would one achieve: Total 5 = 1 pool + 4 extra runners if more events are triggered?

[nocturne1]

This behavior confused me quite a bit as well, as settings seem to be fighting each other. What I've found is

delay_webhook_event = 0
enable_ephemeral_runners = true

runners_maximum_count = 5

# override scaling down
scale_down_schedule_expression = "cron(* * * * ? *)"

# pool usage
pool_runner_owner = "xxxx"
pool_config = [{
  size                = 1
# this change allows an instance to begin startup and be counted by the pool lambda
  schedule_expression = "cron(*/2 * * * ? *)"
}]

enable_job_queued_check = true

# I set this because it will allow pool instances to remain in pool for a full day, if unused.  
# If short, instance will shut down at that time if idle, which will then cause another to be launched. 
# Way too much churn.
minimum_running_time_in_minutes = 1440

With this setup, I have a pool that reliably remains stable, but then grows when requests come in.

I also have seen issues where if runner_name_prefix is defined, there seems to be some confusion between the pool lambda and the scale-up lambda. So I've had to disable that too.