Domain Controller in a VM

[Emulex]

you sync to a physical AD server man. trust me. move the core roles to the VM AD servers and have 2 Virtual and 1 physical or 1 phys and 1 virtual, or best 2 phys and 2 vm that way you have a RTC handling situations.

most people backup during dull times - when everyones offline - you aren't going to tombstone if no changes to the AD are being made. you backup all servers at the same time (veaam for vm's) that way both PHYS and VM backups are time sync'd together - that will give you best chance of survival. imo you want an AD server offsite as well in case you lose everything - it will be alot less trouble if you have one living AD server up all the time - this could be a vpn over comcast business to your home - doesn't have to be fancy but remember some AD traffic can get heavy if you were hosting a ton of exchange domains. just gets costly remember to NEVER run anything else on an AD server - no sql - no exchange - maybe print/light file serving - if you ever demote the server you will roach your exchange/sql as they are unseparable - and the write caching (forced sync) will make it SLOW as a turd. ask me how i know why this is bad inherited a sql server running sharepoint with AD. now i have to migrate sql and sharepoint to another server because the hardware is aging out.

 

[imagoon]

It probably isn't the VMware layer, but it does add enough mystery to make troubleshooting more difficult. Typically, 10:00-11:30am, we'll see indexing slowness within the system...and again between 5pm-7pm when backups start running. That's only on VMware...not the physical mail stores. I know a lot of the backup issues have been resolved in the past week by changing around some Exchange settings...but they were specific to running Exchange Server in a VMware environment from Symantec.

I just think for the cost of a real blade server, it's worth the money to know there's no virtual hardware bottleneck when the system is indexing mail or getting hit with multiple large listserv deliveries/SPAM storms and daily mail storms caused in that 10am hour. (busiest time of the day)

Blade sounds way more expensive to me.* When it comes to HA and failovers, VMware is king of the castle. Again odds are the "vm mystery" is likely not your issue and should be ignored unless there is indication there is a problem at the host level. vSphere will tell you this in the perf and monitoring graphs.

*The blade itself is cheap, support maintenance, power, cooling, licensing, migration etc drive that price up. Sounds silly to spend money to fix what I would guess is "not the issue." If exchange is having issues, check the other VM's on the host. Are they having issues? Is the host CPU bound (easy to check in vSphere, it will even set alerts) did someone over commit the host? Is the SAN IO maxed? Stuff that I am listing points at poor management of the virtual infrastructure. If the VM infrastructure is fine and other vm's on those hosts are running fine, the issue is not VMWare, and it not worth wasting IT resources moving it to a blade. The money is better of being spent "unmystifying" you so you are not throwing stuff at a wall seeing what stuck.