Netdata is actually doing itself a disservice by advertising their totally optional cloud integration service so hard, that you think it actually can’t work without it or provides no value.
From my recent research, if you just install netdata on your node you get the single-node dashboard - and yes the cloud notification is annoying and can’t be turned off.
You can then setup netdata installations on nodes to stream their data, so you can visit node A and also access the dashboards of node B and C. You can also disable data storage on streaming nodes if you want.
The cloud option is free and simply streams the data towards the cloud, all data stays on the machine. What the cloud gives you is unified settings, and alarms as well as some dashboard unifying multiple nodes. This whole cloud thing seems to be free for now (maybe not with this release?), with them wanting to add paid features later on.
After my recent research I still have the conclusion that netdata is just the best thing for monitoring nodes. You get all data over all components, you get alerts that include sending matrix messages and you get streaming between nodes with data retention, such hat the highest data resolution is for 24h and then it gets less detailed, so your storage isn’t overwhelmed. I tried grafana+prometheus+prometheus node exporter first, but netdata just gives you a complete package, a working ZFS integration and alerts out of the box. While I’ve had a hard time getting prometheus/grafana to actually give me information about single-core bottlenecking, as it just summarized the CPU usage. So 10/10 cores on 10% looks the same as 1/10 on 100% utilization. You maybe do want to disable their anonymous analytics on installation.
I think the cloud option makes sense if you want some go-to thing to watch all your nodes, have some unified dashboard and give you some way to just follow alerts directly to the source by linking to their cloud stuff. I think they should provide a way to just say “no, disable this, don’t mention it, thanks”. It also probably makes sense being able to outsource monitoring data, so you at least don’t need a machine for that on top. I really hope netdata cloud-less and netdata with cloud can co-exist in the future and still provide enough funding towards the company.
Edit: typos after the machine died and I salvaged this comment..
The data is available locally in a dashboard. You can also stream to other agents or to other observability platforms. I use it without any cloud service for all my home lab setup. I use the per device dashboards and a shared graphana ui as well.
In fact I think even with the cloud setup all the data still exists on your our infrastructure.
Do you have links how I’d set that up ? I’m especially interested in the grafana dashboard you’re probably using. My experience with finding good node-exporter dashboards, apart from the default one, wasn’t very good.
Great to see such feature rich releases. Netdata is the easiest monitoring solution I’ve used and got value out of.
If I’ve understood their readme correctly, the agent is completely useless without the cloud service, regardless of being open source?
Has anyone forked this to transmit data to something non-proprietary?
Netdata is actually doing itself a disservice by advertising their totally optional cloud integration service so hard, that you think it actually can’t work without it or provides no value.
From my recent research, if you just install netdata on your node you get the single-node dashboard - and yes the cloud notification is annoying and can’t be turned off. You can then setup netdata installations on nodes to stream their data, so you can visit node A and also access the dashboards of node B and C. You can also disable data storage on streaming nodes if you want.
The cloud option is free and simply streams the data towards the cloud, all data stays on the machine. What the cloud gives you is unified settings, and alarms as well as some dashboard unifying multiple nodes. This whole cloud thing seems to be free for now (maybe not with this release?), with them wanting to add paid features later on.
After my recent research I still have the conclusion that netdata is just the best thing for monitoring nodes. You get all data over all components, you get alerts that include sending matrix messages and you get streaming between nodes with data retention, such hat the highest data resolution is for 24h and then it gets less detailed, so your storage isn’t overwhelmed. I tried grafana+prometheus+prometheus node exporter first, but netdata just gives you a complete package, a working ZFS integration and alerts out of the box. While I’ve had a hard time getting prometheus/grafana to actually give me information about single-core bottlenecking, as it just summarized the CPU usage. So 10/10 cores on 10% looks the same as 1/10 on 100% utilization. You maybe do want to disable their anonymous analytics on installation.
I think the cloud option makes sense if you want some go-to thing to watch all your nodes, have some unified dashboard and give you some way to just follow alerts directly to the source by linking to their cloud stuff. I think they should provide a way to just say “no, disable this, don’t mention it, thanks”. It also probably makes sense being able to outsource monitoring data, so you at least don’t need a machine for that on top. I really hope netdata cloud-less and netdata with cloud can co-exist in the future and still provide enough funding towards the company.
Edit: typos after the machine died and I salvaged this comment..
The data is available locally in a dashboard. You can also stream to other agents or to other observability platforms. I use it without any cloud service for all my home lab setup. I use the per device dashboards and a shared graphana ui as well.
In fact I think even with the cloud setup all the data still exists on your our infrastructure.
You can also scrape Netdata with Prometheus. It has more features by default than node_exporter.
Do you have links how I’d set that up ? I’m especially interested in the grafana dashboard you’re probably using. My experience with finding good node-exporter dashboards, apart from the default one, wasn’t very good.
https://learn.netdata.cloud/docs/agent/exporting/prometheus
this misses the crucial part: grafana dashboards that provide the same value