amppackager
amppackager copied to clipboard
ampproject
Prometheus endpoint for metrics
Hey there!
I think a prometheus endpoint would be very handy to have in this service, so we can easily track request/error rates etc without needing to rely on the metrics for the proxied service itself. Once the service is out of development mode it's not easy to operate without having metrics readily available.
Thanks!
Oops, I missed this notification. Agreed, seems like a good idea. Any suggestions for metrics to expose? Off the top of my head:
- # of requests
- failures (by type)
Less important but possibly useful:
- per-request: CPU usage, response byte size, latency (total [request -> response], added [fetch response -> response])
Yeah I think those are pretty much what I had in mind.
Response byte size is definitely important, given that there's a limit on the size of the response that the packager can process, so being able to monitor that would be really useful.
The ability to monitor response codes from upstream as well as latency would also be great.
Thanks for working on this!
Sorry for the delay on this. It's still top-of-mind, but life.
Wanted to document a few other ideas:
- time until cert expires
- time until ocsp expires
- ACME or OCSP fetch errors
@FranGM Hi Fran!
I've implemented a bunch of basic metrics recently per your request. The documentation lists the available metrics. Would you like to try them out and give us some feedback?
Hey there! Sorry about the delay. I might not have enough time to test this in any meaningful way at the moment but I've been going over the linked docs (which are excellent) and this seems exactly what we need. Thanks!