Bandwidth reqs for remote_write / agent mode

13 views
Skip to first unread message

Michael Kogelman

unread,
Nov 30, 2022, 9:51:50 PM11/30/22
to Prometheus Users

I really want to federate our Prometheus across a WAN link to get everything into one pane of glass on our end – that said I’m being asked to quantify the bandwidth usage we may see. Is there any rough guide to the ratio of metrics:bandwidth anywhere?

 

It’s going to be something along the lines of 60 machines, half windows_exporter, half node_exporter if that helps at all.

 

Thanks!

Mike

 

 

--

Michael Kogelman

Principal Systems Administrator, Contract Support

NDP

 

 

 

Brian Candler

unread,
Dec 1, 2022, 4:59:29 AM12/1/22
to Prometheus Users
Obviously it will also depend on your federation scrape interval.

There's no guide that I'm aware of, but it's easy enough to measure: do a federation scrape with curl, and see how many bytes are returned.  (In practice it'll probably be less due to gzipping, but use tcpdump to check that the HTTP exchange is in fact compressed)

You can get an upper bound just by scraping one exporter and multiplying up by the number of hosts: e.g.

# curl -fsS localhost:9100/metrics | grep -v '^#' | wc
   2188    4390  125202

Scrape that once per minute and that's 125KB/min = 1MB/min ~= 17kbps.  But if the HTTP exchange is compressed, I would expect it to be reduced by a factor of at least 3 or 4.
Reply all
Reply to author
Forward
0 new messages