Skip to content

Autoscaling queued

Hello,

I am testing the docker-autoscaler executor with the following configuration:

concurrent = 30 # must match max_instance * capacity per instance

executor = "docker-autoscaler"

plugin = "azure:1.0.0" ## i have also try the 0.3.0 version

capacity_per_instance = 1 # max one job per vm

max_use_count = 1 # use the vm only for one job and destroy it

max_instances = 30 # max vm running

idle_count = 1 # keep 1 vm constantly runnning even with no jobs

idle_time = "20m0s"

My problem is the following: when I run a pipeline with multiple jobs (say 15), I except that the runner will create 15 VMs on my VMSS on azure at the same time, but in fact, one vm at a time is created, when one VM finish the creation, the next one is created, so with 15 jobs (and capacity_per_instance = 1) the last jobs has to wait 14 Vms creation to start (approx. queued 3-4 minutes).

I would like to know if this behavior is excepted or if I have a misconfiguration somewhere.

I already taked a look on Azure side, zero rate limit found and I can create 15 Vms concurrently with Azure cli. I have tried to specify burst and limit in the config too, even if the default seem to be sufficient for my needs.

Thank you for your help.

Edited by Michael Rassu
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information