We currently manage many ESX machines from one server running Windows 2k3 and Virtual Infrastructure Client. If we shutdown the Windows 2k3 machine, this will have no affect on the ESX servers correct? All VMs will still be up and running?
HA is not dependant on vCenter except for the initial configuration. So, if you are shutting down vCenter, maybe just set DRS to 5 stars, and leave HA as is. After all if there is a hardware failure, you still want your guests to failover.
All VMs will still be up and running?
YES. But any automated functions such as HA should be disabled prior to shutting down.
Why would you disable HA when shutting down VirtualCenter?
HA agents on ESX hosts will continue to monitor each other and initiate failovers if necessary even when VC is down.
What will NOT work is DRS and migrations, statistics will not be rolled up in DB, alarms not generated etc. Otherwise nothing critical if VC outage is not planned for a long period.
WBR
Imants
HA agents on ESX hosts will continue to monitor each other and initiate failovers if necessary even when VC is down.
HA is a function of VC also, so it will NOT monitor or initiate a shutdown unless VC is UP,
And I have seen instances that HA fails sometimes or causes VC to hang, so disable it first will make VC maintenance easier..
We have DRS recommendations enabled, but it does not do anything automatically; only recommendations. Do we need to disable the DRS or will it not cause issues?
Seems like disabling DRS and HA is not a big deal for VC maintenance.
Seems like disabling DRS and HA is not a big deal for VC maintenance
It's not. Plus when I shutdown VC, then later it starts back up, some ESX hosts don't respond as quickly as others and then VC will start migrations randomly, so it's a precaution.
Now I'm confused.
Can't find an "availability guide" document for VI3 right now, but in vSphere (4.0 and on) vCenter is only responsible for configuration of HA agents on hosts. Once configured they can run independently and do not suffer from vCenter outages (read: hosts continue listening / sending heartbeats and initiating failovers). This is a normal way to protect vCenter if it's on a VM in small environments.
Could it be that there were such a drastic changes in HA architecture from 3.5 to 4.0!?
WBR
Imants
HA is not dependant on vCenter except for the initial configuration. So, if you are shutting down vCenter, maybe just set DRS to 5 stars, and leave HA as is. After all if there is a hardware failure, you still want your guests to failover.
That's exactly what I'm talking about.
And agreed on DRS automation level, just to avoid misunderstandings - 5 stars is the most Conservative option (I've had cases where people think the opposite way).
WBR
Imants
yes, 5 stars in most conservative.
The shutdown of our VC server did not cause any outages, thanks for the input!
If this downtime is prolonged, say a week or more, are there any further issues involved? All VMs and ESX hosts should continue functioning regardless of where the VC server is running or not. Has anyone seen this to not be the case? Also if after, say a week, we then bring back the VC server are there issues of communication with the ESX hosts since it has not been running for quite some time? I guess I wonder what communcation goes on between VC and the ESX hosts that would be disrupted by the fact that they havent communicated for 1+ weeks.
Thanks!