Clarify EC2 instance size requirements

[seanlinsley]

The differences are that t3.micro has 1 GB of RAM instead of 2, and costs $7.59 a month instead of $15.18.

[msakrejda]

This reverts https://github.com/pganalyze/pganalyze/pull/982 -- is that worth doing? Collector OOMs are not that easy to diagnose, and may include support interactions, creating an onboarding hurdle. It's a difference of $7/month with on-demand pricing.

Maybe instead we should add a separate docs page on sizing your collector instance, and go into more detail there (and link from here)?

[seanlinsley]

Since pganalyze/collector#384 our production collector (which monitors two servers) has reliably stayed below 1 GB. In the image below it peaks at 545 MB.

I'm not sure what benefit a dedicated docs page would bring. Is there really that much to cover?

[lfittl]

Since pganalyze/collector#384 our production collector (which monitors two servers) has reliably stayed below 1 GB. In the image below it peaks at 545 MB.

I'm not sure what benefit a dedicated docs page would bring. Is there really that much to cover?

I'd also be in favor making this a separate page, and continuing to default to the larger instance in the docs.

The cost difference here is negligible for most organizations, and we should have a place anyway where we can clarify that the collector needs more resources (both on the host and the systemd unit file) if you monitor many servers. If someone is truly cost conscious on this level, they could go to that page and decide to go smaller too.