Monitoring Apps

[teknoadmin]

Hello!

I need a monitoring apps for VM usages(cpu, ram, disk, network etc.) What's your suggestion?

Thank you!

 

[UdoB]

While primarily being a Zabbix (https://www.zabbix.com/la/integrations/proxmox) user - and Zabbix works well! - I found a gem somewhere else: https://grafana.com/grafana/dashboards/19119-proxmox-ve-cluster-flux/

My Grafana/Influx-setup followed a random "InfluxDB and Grafana and Proxmox PVE" instruction found at a random website. For this approach you do not need to install additional software on the host. Most other solutions need a daemon of any kind to gather information. For Influx it is integrated by default: https://pve.proxmox.com/pve-docs/pve-admin-guide.html#external_metric_server

In my constrained recognition Zabbix is both more flexible and more complex. (Influx/Grafana users will probably deny this statement.)

LibreNMS (https://www.librenms.org/) was the easiest experience for me: just install SNMPd and a lot of metrics are autodiscovered - but only generic "Linux", nothing really PVE-specific - so it is not optimal.

For all approaches you need a place to run that software. This may be a VM - but monitoring a virtualization hardware makes more sense by running the monitoring software on independent hardware, off-cluster.


Have fun!
Disclaimer: my own monitoring @home is just doing this absolutely fine: running in VMs inside the cluster. Only at work I have a dedicated and reliable server...

 

[teknoadmin]

While primarily being a Zabbix (https://www.zabbix.com/la/integrations/proxmox) user - and Zabbix works well! - I found a gem somewhere else: https://grafana.com/grafana/dashboards/19119-proxmox-ve-cluster-flux/

My Grafana/Influx-setup followed a random "InfluxDB and Grafana and Proxmox PVE" instruction found at a random website. For this approach you do not need to install additional software on the host. Most other solutions need a daemon of any kind to gather information. For Influx it is integrated by default: https://pve.proxmox.com/pve-docs/pve-admin-guide.html#external_metric_server

In my constrained recognition Zabbix is both more flexible and more complex. (Influx/Grafana users will probably deny this statement.)

LibreNMS (https://www.librenms.org/) was the easiest experience for me: just install SNMPd and a lot of metrics are autodiscovered - but only generic "Linux", nothing really PVE-specific - so it is not optimal.

For all approaches you need a place to run that software. This may be a VM - but monitoring a virtualization hardware makes more sense by running the monitoring software on independent hardware, off-cluster.


Have fun!
Disclaimer: my own monitoring @home is just doing this absolutely fine: running in VMs inside the cluster. Only at work I have a dedicated and reliable server...

Is there any chance to set alarm for resource usage of vms? thank you very much sir.

 

[jsterr]

We did a tutorial on howto install influxdb (2) and Grafana with latest software:

https://www.thomas-krenn.com/de/wiki/InfluxDB2_+_Grafana_Konfiguration_eines_Metric-Servers_für_Proxmox_VE
https://www.thomas-krenn.com/de/wiki/InfluxDB2_+_Grafana_Docker_Container_Installation_in_Ubuntu

Its german but commands are universal language so should be easy to understand. Its also uses TCP instead of udp and pve docs are little outdated for influxdb2.

Is there any chance to set alarm for resource usage of vms? thank you very much sir.

Alarms can be set on graphs yes. Do you mean ressource usage per VM or in total light overcommitment? I would go for this dashboard, you can import it (also explained in the above tutorials): https://grafana.com/grafana/dashboards/10048-proxmox/

 

[UdoB]

Is there any chance to set alarm for resource usage of vms?

For Influxdb/Grafana see @jsterr 's answer.

For Zabbix: yes. Zabbix has a komplex alerting system with scheduling by daytime and escalation to different groups of contacts depending on severity and other factors.

For LibreNMS: never tried...

For PVE native, without a separate monitoring solution: no, not as far as I know.

 

[teknoadmin]

For Influxdb/Grafana see @jsterr 's answer.

For Zabbix: yes. Zabbix has a komplex alerting system with scheduling by daytime and escalation to different groups of contacts depending on severity and other factors.

For LibreNMS: never tried...

For PVE native, without a separate monitoring solution: no, not as far as I know.

We did a tutorial on howto install influxdb (2) and Grafana with latest software:

https://www.thomas-krenn.com/de/wiki/InfluxDB2_+_Grafana_Konfiguration_eines_Metric-Servers_für_Proxmox_VE
https://www.thomas-krenn.com/de/wiki/InfluxDB2_+_Grafana_Docker_Container_Installation_in_Ubuntu

Its german but commands are universal language so should be easy to understand. Its also uses TCP instead of udp and pve docs are little outdated for influxdb2.



Alarms can be set on graphs yes. Do you mean ressource usage per VM or in total light overcommitment? I would go for this dashboard, you can import it (also explained in the above tutorials): https://grafana.com/grafana/dashboards/10048-proxmox/

I mean resource usage per VM.
I appreciate for your valuable answers. Thank you very much!