Setting up true up/down alert for ESX Host

[ooajala]

Hello Community,

I am head over heels for this one. I have been trying to setup alert that would give me valid up/down alerting in vROPs for all my ESXi Hosts. I have tried using the Power State and Connection metric in Properties > Runtime, but when I powered down a host to test this, the value for Power State came back as "Unknown" and Connection value was "disconnected".

I thought maybe the Power State value of "unknown" would suffice, BUT I then checked another host that was not powered down but had just lost connection to vCenter and was still pinging, this Power State value was also "Unknown" so figured I can't use this metric.
Has anyone successfully setup an up/down alert in vROPS for ESX Hosts?

Please let me know how you got it done.

[ooajala]

Oh also, just as part of what I have tried, I used the new PING monitoring feature in 8.4, BUT I did notice two things that were an issue:

1. After enabling the PING feature for all hosts for over a week, I come back to check on some hosts under object detail, it doesn't populate the PING details, hence when you check the Metrics, you don't get the Ping Statistics metrics

2. For the Ping Statistics metrics, I am getting additional IPs other than just the management IP such as VLAN IP etc, so I would not be able to use this either for my up/down. I probably would have used the Packet Loss % when it is 100%, BUT there are multiple IPs here and I would be unable to use this metric.

[sennevanlaer]

Hi I think it depends on your definition of down. here, we alert whenever a host gets disconnected or has no running vms. We also monitor gateway ip's to differentiate between a network or host issue asap.

I think that what you want to know is if the vm's are still running - as 'disconnected' already means loss of mgmt. Note that if the host is disconnected from vcenter, you already lost almost all metrics for vsphere host and vm objects. vcenter is crucial to get data from vsphere.

You can setup more specific ping  monitoring using the ping adapter. it would require some manual work to link these to the hosts.

You could also setup a http test for the esx homepage, or monitor vm service endpoints (tcp ports, shares, webservices,...) from an agent running on another vm.

[ooajala]

Actually what I want to alert on is ONLY when the ESX Host is completely off meaning unpingable. I was really trying to avoid using the ping adapter to accomplish this BUT it is practical to do so because if the ESX Host is completely offline, there's no way to retrieve metrics related to the ESX Host for alerting. We will give the ping adapter a go and see what perspires!