Netdata Integration

JohnB

Would anyone else be interested in netdata integration?

Netdata supports external plugins and custom collectors. If the enhance process/containers exposed statistics in the format netdata expects, the netdata dashboard (local and cloud) can collect all enhance statistics.

As an example if the CPU usage spiked, netdata would send an alert email with a link that automatically runs correlations. If the enhance data was included it would be obvious that site xyz has unusually high activity.

Its great to have a simple CPU/Memory usage graph in enhance for the customers but for the server admins if we could output all the statistics for tools like netdata, it makes monitoring the whole picture much easier.

Even just exposing a datastream of performance metrics for each site would allow admins to easily make plugins for their favorite monitoring tools.

https://learn.netdata.cloud/docs/contributing/external-plugins?_gl=1*kyfw7y*_ga*MTc0ODk4NjAwMy4xNjk3NjY0NzAw*_ga_J69Z2JCTFB*MTY5NzY2NDcwMC4xLjEuMTY5NzY2NTM5NS42MC4wLjA.#external-plugins-api

Andreas

Couldn;t this install be enough?
https://www.netdata.cloud/integrations/linux/

JohnB

Andreas That won't integrate with enhance though. You would get the normal metrics which are great but there could be enhance charts / graphs. Things like page views, show all the sites with a different color on a line graph to see which is using the most CPU, most processes etc.

Just installing netdata will only give you docker container names and, depending on the configuration, memory, cpu, network usage. Could be much more with a little support on the enhance side of things.

I got the idea from this post: https://community.enhance.com/d/26-more-monitoring-server-tools-logged I think instead of spending the time to make nice panels for admin monitoring inside of enhance, just integrate with a much more robust monitoring platform.

AdamM

JohnB
+1

Are you now just using NetData on your servers without configuration?
I'm thinking whether to give it a try, do some testing.

Oldschool

I would love to see this, or at least more monitoring of servers for admins.

JohnB

AdamM I use netdata on all of my servers. Create a cloud account (completely free) and ran the agent installer. Its really seamless.

HOWEVER... you should know that default install will expose netdata to the public with no password on port 19999. Its read only sensor data but would be useful data in DDOS attacks and some types of hacking.

What I do is just block port 19999 because netdata pushes data to the cloud account. Their cloud servers don't poll, so blocking 19999 doesn't affect anything.

If you want just a local install (https://learn.netdata.cloud/docs/installing/one-line-installer-for-all-linux-systems) you can change one line in the config file to bind to localhost only and then setup a reverse proxy with a password. You can find instructions in a github issue.

AdamM

JohnB

Thank you for your detailed answer.
I just created an account and installed the agent, it looks very good, it provides a huge amount of statistics. I can even see that there is data of resource usage by docker and containers. Great!

Of course, I have blocked all ports in the firewall that are not needed for Enhance.

necrom

We'll be working on some bash scripts to recurse the /sys/fs/cgroup/websites/ files and build a json (if anyones done this already that'd be really handy and work for any monitoring system). We'll then be using zabbix discovery on them and visualising in Grafana.

EDIT: We'll release our scripts and our zabbix templates once they've matured.

JohnB

AdamM Np. Oh and there are lots of plugins you can find. That's what I was thinking when I mentioned enhance integration. Just a plugin that grabs enhance stats.

To @necrom point it would seem this plugin might be what you want if use netdata for monitoring: https://github.com/netdata/netdata/blob/master/collectors/cgroups.plugin/README.md

necrom

JohnB That's useful cheers, will look to convert into zabbix (we prefer self-hosting our monitoring without any limits). Or I need to work out why our existing docker cgroup monitors don't include enhance (https://www.zabbix.com/integrations/docker).

JohnB

necrom (we prefer self-hosting our monitoring without any limits)

The fewer external companies a hosting provider depends on the better right lol? For me, I found setting up Grafana to be way more work than I want for just monitoring but I've seen some amazing dashboards and custom capability lol.

But just FYI you CAN self host netdata and selfhost notifications. I think there is a plugin that can do correlations without their cloud as well.

necrom

JohnB The fewer external companies a hosting provider depends on the better right lol?

Hah in a perfect world maybe but some things need special attention/software, our zabbix has 49928 items and 352.52 new values per second which I dread to think how much would be in a cloud/SaaS. Our grafana pulls from lots of different places that would be hard to visualise in the tools themselves (mostly zabbix, whmcs and multiple JSON sources like endoflife.date).

But just FYI you CAN self host netdata and selfhost notifications

Interesting, I took the 5 node limit on community edition to mean agents/hosts (and we have 17 zabbix proxies), will take another look.

JohnB

necrom node limit on community edition

Ahh I wasn't aware of that. The times I've self hosted it, it was just a few nodes. That said, it would be worth checking if there is a reliable way to increase that limit.

necrom our zabbix has 49928 items and 352.52 new values per second which I dread to think how much would be in a cloud/SaaS.

Surely the Saas economical plan would be "max 300 values/second" and the next plan jumps to 3000 values/second lol... Oh more than 3 days of history? "Call for pricing"

Rich

+1 for sure, I've got some horrible load spikes on my dedicated server, that's also the CP and I am using it to try and identify what's going on. It's client PHP processes, but without Netdata I wouldn't of been able to of known unless catching it realtime.

Not had the time to iron that out, but I do have Netdata on the nodes, and yes docker integration would be great. Likewise a UI page to any self-hosted instance(s). (basic integration module in enhance maybe)

Jordan

I don't know if anyone has done anything in-regards to monitoring nginx/litespeed/ols, mysql or redis at all. But I'll be working on the scripts to setup netdata and the specific services.

We might need Enhance to expose some sockets for server status for nginx/litespeed/ols, and PHP-FPM.

FabioP

+1 for me.