When I create every new VM for general purpose Apps (often not known HW geometry previously), I'm ask myself - how many vCPU should I create and how it must be balanced - more to sockets or to cores? And would scaling vCPU counts may take worse for VM working? Finaly, _what happening_ w/ guest's OS and apps - when I do it. I think it is not exact, like when we set second/third-fourth CPU to HW server in empty slot, or change CPU's (physical) w 6 cores to modern 8/12 etc.
So I have couple questions set:
Firstly:
so f.e. (from one hands) should I create 2 vCPU -s better 2 sockets and 1 core per socket, or 1 socket w/ 2 cores. Or maybe better way make all of them in full set.
From other hands, I was told on courses VMFT 5.5 that many vCPUs in rare case better than few of them (2-4), and more in count of sets it maybe worse. So better way make 1-2 vCPU.
Secondly:
is well known OS-s (f.e. Windows Server 2008-2016, linux ubuntu /redhat 5-7, freebsd) sense vCPU's on setup, or works worse when I scale vCPU-s instead when i give them final set of vCPU? So maybe I should create VM w/ 4 vCPU instead 1 or 2 as minimum. And, secondly, is a different w/ starting 1 vCPU vs 2 vCPU.
Also to Apps, is common purposed Apps sense that scaling? I mean such apps as windows services (WSUS, AD) or mail systems (windows exchange or linux sendmail/postfix) or webservices (Firefront, apache/tomcat). Also and more interest RDBMS apps: MS SQL, Oracle database (standalone/RAC), MySQL.
P.s.: I did scale few times and didnt noticed smth especial w/ VM guest and apps internal. I didnt noted those events and didnt inspect VM health after that.
But a question is ...
TY.