Ask HN: Best environmental monitoring front end for large lab, small datacenter?
We have about 3000 physical servers in our R&D lab. We have various environmental sensors (server temps, rack temps, room temp, humidity, HVAC statuses, etc) already accessible behind disparate portals.
I'm moving all these to one frontend. Are there some solutions for this that people here like? I'd love to be able to share all the dashboards with the rest of my team, send out emails automatically when things break, and also send emails when they appear fixed.
Obviously it's not too hard to do this from scratch, but figured it might make sense not to reinvent the wheel if there's something excellent available.
6 comments
[ 2.9 ms ] story [ 24.5 ms ] threadA common monitoring stack is Telegraf/InfluxDB or Prometheus for monitoring and storage. Grafana for visualization.
As for systems/protocols, all I know at the moment is that some of it currently uses Zabbix. It's multi-departmental so fact-finding goes slowly. I'm an industrial IoT guy at heart, so I'm also eager to determine the current topology.
I've used some proprietary solutions for things like this in the past (OSIsoft PI System, no relation to the Open Source Initiative "OSI")...and Sunbird DCIM looks fantastic. Both are probably a contracting company's dream for earning huge contracts to configure and maintain, and while I'd love to have the features Sunbird advertises, it doesn't feel like the right solution during a cost-cutting year. It also feels like it would be more difficult for the team to maintain.
In addition, my organization has been willing to let employees use their 9-5 work hours to contribute to open source when bug fixes or new features were needed for software stacks we depend on. So it feels good to see some well-developed open source options that we may be able to contribute to down the road.
Zabbix can integrate with InfluxDB and I beleive there is currently a Telegraf Input Plugin for Zabbix under development. Grafana can also be used to query Zabbix content directly.
I have extensive expierence with the traditional Historians like OSI Pi Server. You can actually visualize data from OSI Pi in Grafana! The Open Source and COTS produts have their own specific use cases. Much of it comes down to will the IT or OT groups be maintaining the system and what kind of support do they require.
That's great that your organization supports open source!
If you want a fancy easy to use UI you can consider getting the data into something like DataDog. But it’s expensive.
For lower budget, consider Prometheus and Grafana. You will need to write glue code to get the the data points wired to prometheus.
For a scalable backend consider Grafana cloud, or you can run Cortex or Thanos yourself.
What resolution do you want? That is how frequently do you want to save a datapoint?