Host and VM Disk Usage Alarms

[alhamad]

I conifigured host and vm disk usage alarms. I constantly noticed that those alarms are trigerred at least twice a day. My setup consists of two esx 3.0.2 hosts in DRS and HA cluster connecting to a SAN with redundant fabric switch over redundandt 4 Gb port.

How to make sure that there are no IO bottlenicks? Also, the vm alarm does not show which one has the trigger. How to solve this?

[RParker]

What do you mean "the VM alarm does not show which one has the trigger"? do you mean you created an alarm, and you combined more than 1 trigger for the alarm?

Also what does your alarm do exactly. Are you setting a threshold to determine when IO or disk requests goes over a certain number?

Normally you monitor IO bottlenecks at the SAN and watch the Fibre traffic, what kind of SAN do you have?

And you have only 2 hosts, but you use HA? You are aware that HA reserves the equivalent of 1 ESX host in a cluster, right? So you have 2 hosts, you are basically only getting 50% of your usage.

I don't think you want to do this with only 2 hosts, that's a waste. At least get 1 more server for HA.

[alhamad]

I set an alarm to notify me by email if the host disk usage exceeds 75 or 90%. I also set another alarm to notify me if a vm disk usage exceeds 75 or 90%. Examples of alarms received for both host and vm are below:

Target: host-name

Old Status: Green

New Status: Red

Current value:

Host Disk Usage - (Metric Disk Usage (Average/Rate) = 94%)

Target: host-name

Old Status: Green

New Status: Red

Current value:

Virtual Machine Disk Usage - (Metric Disk Usage (Average/Rate) = 101%)

I'm using HP EVA 4000 SAN.

As for the two hosts, we are planning to expand that's why we have HA up and running. However, I do not understand why you have a point on HA running only on two hosts. This way if one host fails, the other one will automatically restart all VMs running on the other system.

[alhamad]

I set an alarm to notify me by email if the host disk usage exceeds 75

or 90%. I also set another alarm to notify me if a vm disk usage exceeds

75 or 90%.

I'm using HP EVA 4000 SAN.

As for the two hosts, we are planning to expand that's why we have HA up

and running. However, I do not understand why you have a point on HA

running only on two hosts. This way if one host fails, the other one

will automatically restart all VMs running on the other system.

[sheetsb]

The point regarding HA with two nodes is that you reduce your capacity by half. HA reserves enough CPU and memory capacity to make sure that all the VMs in the cluster can run if you lose hosts. If you only have two hosts in your cluster you are reserving one host's capacity which allows you to run only half the number of VMs you might otherwise. By adding more hosts you spread the reserved portion of CPU and memory across all of them and thereby increase the number of VMs you can keep running. In theory you are better off to have the maximum number of hosts allowed in a cluster (16 I believe for HA) so that you can maximize your capacity.