[14:31:52] I have some noob questions about dashboards, if I may. I imported a Ceph dashboard to https://grafana.wikimedia.org/goto/Hd-3IBZHg?orgId=1 And it Mostly Works. But it'd be nice to restrict data source to the only 2 useful ones (eqiad prometheus/ops and codfw prometheus/ops) ; understand what "Interval" is doing (cf refresh rate & time picker) ; and maybe take advantage of cluster info (currently there's only 1 such cluster per DC, but [14:31:52] we expect that may grow and/or we'll start collecting metrics from other extant clusters), which I'm guessing would involve massaging the JSON before import? [14:32:19] dashboard JSON comes from https://github.com/ceph/ceph/tree/reef/monitoring/ceph-mixin/dashboards_out and I'll probably want to import some more of these, so it'd be good to get it right now. [14:32:35] [I assume if I want to change the default refresh time, I just log in, set it to what I want and hit save?] [14:32:56] ^-- if all that is too much waffle for IRC and you'd rather a ticket, LMK [14:41:00] Emperor: in the dashboard Settings->Variables, the datasource variable can be limited with an `instance name filter` using regex [14:43:29] `Interval` is the evaluated interval sent to the datasource for how it calculates and renders the graph. (usually time range/max data points) [14:44:29] The dashboard is currently not editable in grafana, so I'm not sure what you need to do to extend it to use cluster info. [14:45:27] I made it editable (and then put it back again) to make the datasource change (thank you for that, worked nicely) [14:45:36] feel free to make editable to have a poke round :) [14:48:55] ok, to do a per-cluster split you'll want a variable with each option (probably based on a query) and each prometheus expression will need a filter to select the cluster using that (think something like `ceph_health_status{cluster="$cluster"}`. [14:51:45] Interval> the other dashboard I use a lot (Swift) does have that adjustable at all; does it default to "auto" if not explicitly defined? [and is that sensible, i.e. should I be thinking about removing the option to tweak it from the dashboard] [14:54:06] auto is usually fine. IMO, it's really only useful when you must manage rate intervals for proper rendering or are querying over very long periods (7d or more) [15:01:24] so maybe set a variable as label_values(ceph_cluster_total_bytes{job=cephadm},cluster) (to restrict us to cephadm clusters) and then use that? [15:03:26] then it's "just" a matter of some JSON editing [15:16:14] that sounds about right :) [15:21:00] thanks. I think that also means I can procrastinate doing that sort of variable work for now if needs be, since we probably won't have another cephadm cluster in the very near future