Replies: 2 comments
-
Hi @bluecmd, thanks for initiating the discussion. I wouldn't characterise what IXP Manager does now as doing its own data collection - in fact all of the graphing used by IXP Manager is external:
But that nit aside, I fully take your point and, in fact, when we designed Grapher (IXP Manager's graphing engine) it was purposely with the idea of allowing other backends and my mind then was firmly on the likes of Carbon (Graphite), OpenTSDB, Grafana, etc. I'm particularly interested in giving the peer to peer elements an overhaul so we can provide more useful information to our members and this requires a proper time-series database. The past few months was all about IXP Manager v6. We now have time to start looking at other elements and I have three specific areas to turn to:
|
Beta Was this translation helpful? Give feedback.
-
We have now implemented Grafana graphs inside our IXP Manager. It looks like this: The graph is fully interactive as you'd expect from a Grafana graph. The way we implemented this is by creating a Grafana dashboard that allows drill-down by relevant metrics (e.g. switch and port) and then embed those graphs via iframes. It works surprisingly well on both desktop and mobile - and the look & feel is quite an upgrade if I may say so myself. The code is by no means pretty, but it works for us. Happy to share and discuss if this would be something that IXP Manager would like to adopt in mainline. |
Beta Was this translation helpful? Give feedback.
-
Hello!
I was wondering if there are any community thoughts around where IXP-Manager might move towards in the future w.r.t. metrics and monitoring.
Right now it seems as a new user that IXP-Manager does its own data collection (SNMP, sFlow, etc.) - while in the areas I work in the trend is more around centralizing the metrics to allow for ad-hoc querying (like Prometheus + Grafana).
I think it would be really cool to see IXP-Manager + Grafana with a Prometheus backend as an alternative to doing any monitoring / metrics collection itself. One positive thing for this is that I wouldn't be forced to rely on SNMP or sFlow for example - if I had other ways of getting that information, I could surface those metrics in my own way to my Prometheus server and IXP-Manager wouldn't need to care.
Thoughts?
Beta Was this translation helpful? Give feedback.
All reactions