Netdata plugin for XCP-ng
-
For those who don't know, Netdata is a kind of "agent" that gather metrics and expose a nice dashboard in live. See a live demo here: http://95.216.43.14:19999
We asked Netdata devs to add Xen related metrics, and guess what, they did it!
It's still in a dev branch right now, but you can see the Xen related chart in the first link given.
What's great with Netdata is the fact you can "stream" to a central Netdata host ("push data" model, no need to declare anything in your central monitoring host) and also connect this central node to a Prometheus/Grafana to get long term metrics.
For those who want to test right now with Xen metrics:
This is NOT recommended in production until we got a binary package, this could BREAK your host yum install yajl-devel xen-dom0-libs-devel libuuid-devel zlib-devel zlib automake autoconf git make gcc --enablerepo='base' git clone --branch xenstat https://github.com/vlvkobal/netdata.git cd netdata ./netdata-installer.sh
-
git clone --branch xenstat https://github.com/vlvkobal/netdata.git Cloning into 'netdata'... fatal: Remote branch xenstat not found in upstream origin Unexpected end of command stream
-
@kevdog this branch is merged since a while
You can use masterif you like. Also, we are prepping a binary package, if you prefer to wait.
-
@olivierlambert Excited to see the binary package!
-
I just went ahead and compiled from source -- probably not recommended since I had to add a lot of packages from repository but it doesn't seem like right now it broke anything. I did have to modify the firewall and /etc/netdata/netdata.conf file (for those attempting to do something possibly similar). Although looking at the netdata tables is really cool -- I'm not sure if its getting me any closer to figuring out why my dom0 cpu utilization is so high however (https://xcp-ng.org/forum/topic/1763/xcp-cpu-performance-spikes-greater-than-90-a-few-times-per-day-making-all-vms-unusable/3)
-
@olivierlambert will you announce somewhere when the binary is ready?
-
Yes, we will. We'll see that when @stormi is back from holidays (end of August). I'd like to emphasis about the cost of doing this package, because the RPM done by Netdata is far from being viable (see how big is the issue we reported: https://github.com/netdata/netdata/issues/6619). I hope we'll find a way to be able to cover this kind of work with our pro support!
