Background

We have customers who are very interested in measuring and optimizing the time it takes for an engineer to create or start a workspace. When a workspace is starting, two primary steps are kicked off:

• provisioner job (terraform apply to create infrastructure)
  • current metric: coderd_provisionerd_job_timings_seconds
• waiting for agent to download and connect on workspace
  • current metric: coderd_agents_up (kinda, it's not a timing but it changes from 0 to 1)
• waiting for startup script to complete
  • current metric: none

We have coderd_provisionerd_job_timings_seconds for jobs but not for how long it takes for a startup script.

Proposal

• Add metric for how long it takes for startup script
• Add metric for how long it takes for agent to connect
• Add overall metric for workspace start time