Add prometheus exporter. #139
Conversation
promunifi/clients.go
Outdated
| labels := []string{"name", "mac", "site_name", "gw_mac", "gw_name", "sw_mac", "sw_name", "vlan", "ip", "oui", "network", "sw_port", "wired"} | ||
| labelWireless := append([]string{"ap_mac", "ap_name", "radio_name", "radio", "radio_proto", "channel", "essid", "bssid", "radio_desc"}, labels...) | ||
| return &uclient{ | ||
| Anomalies: prometheus.NewDesc(ns+"anomalies_total", "Client Anomalies", labelWireless, nil), |
There was a problem hiding this comment.
It looks like this isn't a counter.
| Anomalies: prometheus.NewDesc(ns+"anomalies_total", "Client Anomalies", labelWireless, nil), | |
| Anomalies: prometheus.NewDesc(ns+"anomalies", "Client Anomalies", labelWireless, nil), |
There was a problem hiding this comment.
it's a counter that resets. very odd. lets have a comment chain here. unless you're on discord?
I just uploaded a new clients dashboard. Please grab it again. It'll be easier if you tell me what changes you make so i can make them on my dash. I'm editing a few things at once.....
There was a problem hiding this comment.
I also just realized some labels are screwed up in UAP :( gonna have to fix that up.
There was a problem hiding this comment.
Ok, I wonder what triggers the resets. It doesn't seem like a normal counter to me.
My changes:
[$__interval]for rate, and setting a min step to 1m. This makes for nice zooming.- Graph null as null, not connected
- Graph fill to 0
- Multiply byte rates by 8 to make them bps, since network people like to see Mbps not MBps.
There was a problem hiding this comment.
oh, also, you're probably going to want to run 1m interval. newer unifi controller doesn't update every 30s...
There was a problem hiding this comment.
All of those options sound good except the part where we do math. I'd like to avoid that and just present the data as received from the controller. I'm trying to keep this as similar to the existing dashboard as possible. Let me see if I can find all those thingies you tweaked..
There was a problem hiding this comment.
This doesn't work? it just produces no data points. rate(unifipoller_client_transmit_bytes_total[$__interval]) but using [5m] does. This is the range vector it uses to find two points to calculate the rate. 5m seems reasonable and will overcome any missing points because the poller went down for a few minutes. no?
There was a problem hiding this comment.
You need to set a "min step" in order to make sure $__interval doesn't get smaller than your scrape interval.
It's too bad the Unifi data updates are so slow. I'm used to having much more data in Promehteus. 😁 I'm polling the APs ever 15s via SNMP right now.
There was a problem hiding this comment.
I'm not sure, that really doesn't sounds like the right thing to put into the range vector. Even 10 or 15m seems more reasonable than locking it to the artificial interval you're setting at 2m. 2m doesn't provide the best resolution: 1m does, and if you set the min step to 1m and try to use $__interval as the rate range vector it doesn't work because it can't find two points in 1m.
There was a problem hiding this comment.
Correct, if your scrape interval is 1m, 2m in the "min step" is the minimum setting you can do. The $__interval is in Grafana for exactly the use case of auto-scaling the range vector based on the graph. The min step is the minimum, not a lock on the step.
If the graph is scaled out to a week, $__interval will be adjusted accordingly, and min step will no longer apply.
|
I've setup the latest beta dashboard here. I made some changes to a few things. |
Almost complete. Feedback required. Closes #88.
The initial contribution will include support for storing metrics in Prometheus. I'm not updating any repo documentation beyond the few config file changes and manual updates you see here. The feature will be mostly silent, but available for advanced users and beta testers.
Later I will build some simple dashboards to get people started. We can iterate on them in time. Once I'm satisfied this implementation doesn't suck and we have at least 3 or 4 dashboards with the basic data displayed, I'll rebrand the product as "including prometheus support." This will take time as there are a lot of things that need to be reworded.
Completed:
Todo:
Clientsis done.