Expanding current environment. Need assistance

[vxaxv17]

We have been running our current setup for about 2 years and are looking to expand it.  Are there any documents that descibe how this is best done?  I've been unable to find much regarding this.  Specfically, we have 5 ESXi hosts in the DC which are running about 50 VMs right now.  We would like to add 2 additional ESXi hosts for a total of 7.

We also are out of ports in our iscsi storage switches.  We currently have 2 Dell 24port switches stacked and would need to move to 48 ports to accommodate the additional hosts.

My questions are basically:

1. Being that the current hosts are 2+ years old and the new ones would not be the same model, how is this typically handled with regard to EVC mode and specs (# of cores/amount of ram)?  If I have to downgrade the new servers to match the specs of the old ones, how does one ever actually upgrade to newer hardware without replacing all hosts at once?  I fail to understand how this would work.

2. Since there are many production VMs running that cannot be taken down, how can I go about replacing the iscsi switches without interrupting the VMs?  All VMs have redundant iscsi paths through both switches in the stack so can i just build the new 48port switch stack and move 1 of the paths over to new switches, make sure those are up and then move the other cables over?  Im not sure how the VM would behave in this scenario with having redundant paths split up like this.

Again if there are some resources I can look at that answer or discuss these items that is great but I have not been very sucessful researching this on my own.

Thanks.

[Alistar]

Hello there,

to answer your questions:

1. The best thing you can do now is to match your hosts' previous specs as closely as possible - this goes for the CPU frequency/amount of cores and the amount of RAM. Then you set your cluster's EVC mode to the oldest processor architecture you have in your cluster - you can do this at the ESXi hosts' run-time. You can always go down in the EVC specs, but to enhance the processor architecture in EVC, up the VM needs to be shut down and powered up to reinitialize which processor masks are exposed. So in my opinion you are all right there. Do not make too much hassle if you can't find the exact same CPU make / frequency - the DRS does a good job even with imbalanced resources, but it is for the best that you keep everything streamlined. If you could post your old/new config we could help you out some more.

Also in upgrading the environment the raw horsepower of Core Count, Clock and RAM size/speed matters much more than current CPU instruction set. You get no degradation in performance and I guess that the applications that use the newest instructions sets are very scarce.

2. You would just keep the path to your switch that is still in place active, set the policy to Most Recently Used, force the current paths on the in-place switch to be used, disable the path to the switch that is being replaced, then exchange the switch, re-cable, re-enable the path policy to Round Robin (if you have), and there you go :smileyhappy:. Do the same for the second switch and profit.

Hope it helped a little.