I've noticed this before, but again just now. I had a 2 vCPU VM suddenly, for about 2-3 minutes of time, take up around 90% of the CPU on a 2 pCPU (2 core) host (normal for this guest, which is a BOXI jobserver). By the time DRS started migrating the 7 or 8 other guests, the 2 vCPU hog was already done and within a couple minutes VM's were being migrated back to the host.
Is it documented anywhere how long DRS takes to make a decision to migrate?
Resource utilization from the perspective of the SQL server and the virtualcenter server might as well have been non-existent so it wasn't some delay in processing.
FWIW I have my DRS set to Fully Automated and aggression 4 of 5.
I tried to find this information myself, but you may have the DRS set to aggressively, try making it more conservative.. My guess is you have it on 1 or 2 stars, I would go with 3 or 4 stars, and only make it more aggressive if you start seeing your hosts be too overloaded.
I tried to find this information myself, but you may have the DRS set to aggressively, try making it more conservative.. My guess is you have it on 1 or 2 stars, I would go with 3 or 4 stars, and only make it more aggressive if you start seeing your hosts be too overloaded.
Good point, I didn't even consider that as a potential factor. I did have it set to two stars. I've up'd that now. Will see how it behaves.
Thanks.