Watcher to check if the VM is down

Hi,

We are using metricbeat/ELK for monitoring our kubernetes (EKS) cluster. I would like to create an alert if the node is down. May I know how to do it?
Because when the VM is down, the metricbeat daemonset can not send the data and how can I create alert that the node is down.

@Ayush_Agrahari

What version of ELK are you using?

There's a couple different philosophies on this.

  1. Alert when the telemetry (metricbeat doc count = 0 for some time interval group by host) is stopped sending. This tells you that metricbeat is not shipping telemetry but does is not necessarily mean the VM is down, perhaps a good indicator but not necessarily.

  2. Another way to do this is use heartbeat and up time to ping to be VM externally to see if it is up and on the network.

2 related but slightly different approaches.

I asked about the version because the newer versions with the kibana alerting have an easy way alert when the telemetry stops

Thanks @stephenb. I am using 7.9.0.

Ping and ssh are disabled on the servers. So may be I need to go with approach-1.

Yup understand there is also ICMP for heartbeat, but understand

So method 1 you can do in watcher OR if you looks at the new Kibana alerting framework you just select an option to get alerted when there is no data

This is a Metric Threshold

Thanks @stephenb.

1 Like