What EC2 Instance Types does everyone use?

[Dave Cadwallader]

I recently asked about sizing recommendations and got some good guidelines.  

But I'm curious for those who use EC2, what specific instance types do you use for Prometheus servers?  And how many targets are you scraping?  

Thanks!

[Ben Kochie]

Not ec2, but GitLab is scraping about 650k metrics, 650 targets, 40k samples/second.

Our instances have 8 CPUs and 28G of memory.  Prometheus 1.x needs 6 cores, and 18G of memory.  2.0 is doing better, but we're still tuning things.

I also just discovered that we do have some recording rules that are chewing up a lot of memory in heap allocations.

--
You received this message because you are subscribed to the Google Groups "Prometheus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to prometheus-users+unsubscribe@googlegroups.com.
To post to this group, send email to prometheus-users@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/f5429cfa-4316-4ef9-8087-3f4de20d6427%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.