What is the difference between these three types of metrics for a virtual machine?
1. Disk
2. Virtual Disk
3. Datastore
I want to see the average total latency for a single virtual machine including all of its virtual disks, should I use the "datastore" metric or should I use the "disk" or "virtual disk" metrics and sum them?
If I use "datastore" will that include informaiton from other virtual machines that have disks on the same datastore?
thanks!
So a quick one.. if you want more detail let me know
The way I typically troubleshoot latency issues is by starting with the VirtualDisk metrics... this gives me a good indication of what the VM is experiencing... once I have looked at the aggregate and determined the VM is suffering I will then break it down to the individual SCSI id's of the disks and determine if it's a problem for only a specific disk on a dedicated datastore etc...
Using datastore latency is not always a good indication as you can have some faulty hosts but because the datastore metric is an aggregate of all ESX hosts latency to the datastore it can actually mask a real performance issues....
For example if you have 10 hosts connected to a datastore and 9 have low latency but 1 has high latency the latency is averaged out for that datastore to look like the performance is fine when in reality 1 host is suffering along with its VM's...
When I find high latency on a VM and then look at the datastore latency and it's low I suspect a problem with the paths to the storage for that host...so I will then look at the HBA latency for the host to break it down to a path.
Hope that helps.
Cheers
So a quick one.. if you want more detail let me know
The way I typically troubleshoot latency issues is by starting with the VirtualDisk metrics... this gives me a good indication of what the VM is experiencing... once I have looked at the aggregate and determined the VM is suffering I will then break it down to the individual SCSI id's of the disks and determine if it's a problem for only a specific disk on a dedicated datastore etc...
Using datastore latency is not always a good indication as you can have some faulty hosts but because the datastore metric is an aggregate of all ESX hosts latency to the datastore it can actually mask a real performance issues....
For example if you have 10 hosts connected to a datastore and 9 have low latency but 1 has high latency the latency is averaged out for that datastore to look like the performance is fine when in reality 1 host is suffering along with its VM's...
When I find high latency on a VM and then look at the datastore latency and it's low I suspect a problem with the paths to the storage for that host...so I will then look at the HBA latency for the host to break it down to a path.
Hope that helps.
Cheers
If you have not already upgrade to 6.4 as there are some new dashboards that are good and may give you what you need
OK great - thanks again!