esitrepxe
asked on
Vsphere 4.1 HA Cluster
I have 4 ESX4i servers with shared iSCSI storage.
No HA cluster has been setup as yet & he servers are in production mode.
Can anyone see any problems with creating a HA cluster in the live enviroment?
No HA cluster has been setup as yet & he servers are in production mode.
Can anyone see any problems with creating a HA cluster in the live enviroment?
The availability guide should help you if needed:
http://www.vmware.com/pdf/vsphere4/r41/vsp_41_availability.pdf
~coolsport00
http://www.vmware.com/pdf/vsphere4/r41/vsp_41_availability.pdf
~coolsport00
It could. But if you do some testing you should be fine.
I would make sure you can ping each ESX server from the ESX servers. Also make sure the ESX servers can ping servers that are located on other ESX servers.
Since you have vSphere Server up it seems these steps might be unnecessary but they are double assurance that there will be no issues.
I would make sure you can ping each ESX server from the ESX servers. Also make sure the ESX servers can ping servers that are located on other ESX servers.
Since you have vSphere Server up it seems these steps might be unnecessary but they are double assurance that there will be no issues.
The main issue that users encounter is not 'downtime', but errors...specifically DNS errors. You need to make sure you have DNS set up properly. Here is a VMware KB that discusses 'troubleshooting HA':
http://kb.vmware.com/kb/1001596
~coolsport00
http://kb.vmware.com/kb/1001596
~coolsport00
ASKER
Thanks for the info
Any thoughts on any best practice changes I should make to the default settings a cluster creates?
Any thoughts on any best practice changes I should make to the default settings a cluster creates?
Hmm...well, the Guide I provided is what is suggested by VMware, of course, but any setting changes from the defaults would really be determined by your org requirements.
~coolsport00
~coolsport00
I would turn on VMware DRS as well as long as VMotion is setup correctly. This will move servers when an ESX is overloaded.
A nice feature is the Power Management (it will move VM;s and power off unused ESX's) but with iSCSI you have to watch out. VMware does not take into play the iSCSI and you can overload your iSCSI bandwidth,
A nice feature is the Power Management (it will move VM;s and power off unused ESX's) but with iSCSI you have to watch out. VMware does not take into play the iSCSI and you can overload your iSCSI bandwidth,
Do all of the ESX hosts have the same processors?
Hi,
Adding HA in live environment it will not effect anything,but if you make sure that all requirements of HA should be fulfilled,there is a document for this http://www.vmware.com/pdf/vsphere4/r41/vsp_41_availability.pdf
Adding HA in live environment it will not effect anything,but if you make sure that all requirements of HA should be fulfilled,there is a document for this http://www.vmware.com/pdf/vsphere4/r41/vsp_41_availability.pdf
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Hi esitrepxe,
The default HA policy is set to "Host failures cluster tolerates" which is very conservative. This will affect the available (not reserved) resource capacity of your cluster. I believe that setting your policy to reserve a percentage of cluster resources is more appropriate in most situations; using the percentage policy make sure you also enable DRS since it takes care of resource fragmentation in case HA is initiated.
Also in order to avoid false isolation response it's a good practice to have a redundant service console (ESX)/management network (ESXi) since this is the network HA uses to exchange heartbeats. These are also the IPs other hosts will try to ping before declaring a host failed/isolated.
The default HA policy is set to "Host failures cluster tolerates" which is very conservative. This will affect the available (not reserved) resource capacity of your cluster. I believe that setting your policy to reserve a percentage of cluster resources is more appropriate in most situations; using the percentage policy make sure you also enable DRS since it takes care of resource fragmentation in case HA is initiated.
Also in order to avoid false isolation response it's a good practice to have a redundant service console (ESX)/management network (ESXi) since this is the network HA uses to exchange heartbeats. These are also the IPs other hosts will try to ping before declaring a host failed/isolated.
ASKER
Not a concise answer, pointing to standard doc's i had already looked at. I was asking for real world answers not just to be pointed to vmware manuals
"exitrepxe", your initial post didn't ask for 'real world answer', but regardless, the Guides provide the answer for that as those *are* best practices and what most org's implement. Regardless, being that I'm one who uses vSphere and an HA/DRS cluster, to answer that question specifically, yes, you obviously can create a cluster in a 'live' environment. Since you didn't explicitly ask for real world scenario, your grade of "C" shouldn't be warranted. When creating posts, make sure you ask specifically what you're wanting. That will help us provide the best answer to your question.
Regards,
~coolsport00
Regards,
~coolsport00
~coolsport00