I started working a while ago managing some F5 hosts and guests and I encounter a strange situation.
During an upgrade from v13 to v15 of a 1 core per slot vcmp guest I ran into a memory issue and I was unable to upgrade this guest.
I was suggested to reconfigure the cores in this vcmp guest to get more memory.
My current setup is the following:
Viprion Host with 4 slots with 20 cores each. (We have 2 chassis)
18 Of the cores in every slot are used. So I have 2 vcmp guests with one core per slot.
I was suggested to reconfigure the vcmp guests using 2 cores in 2 slots so I can get double the memory, moving from 1x4 to 2x2.
I read all the manuals regarding resource allocation but I can't seem to understand why will this perform better (aside from the fact that it will have more memory).
Is it better to have more cores assigned per slot rather than using all 4 slots? Will it handle the same amount of traffic the same way?
F5 recommends to assign 2/4/6/8 and so on cores never 1/3/5 etc. so the best practices were not followed also 1x4 or 2x2 will have the same memory as memory is given for 1 core, so to have more memory you may need to do 2x4. Please see https://support.f5.com/csp/article/K14218 . Better assign half the cores from 1 slot and the other half from other as if blade fails and you need RMA not the entire vCMP quest will crash, it will still work but with half the CPU and memory, so better use 2x4 than 1x8. See below to get the idea:
As when the vCMP quest is on two blades it is just a plug and play, you remove the faulty blade and add the new one and that is it.