*may or may not be* enough - a huge amount depends on your workloads. As ever, watching the available memory and how much CPU is in use in "top" will give you a better idea of how your infrastructure is behaving, as will watching the server charts in virt-manager if it's available. For example, for some of our workloads, we can overcommit memory by a factor of two (so run nearly 200G of virtual server RAM on a 96G machine) and overcommit CPU by nearly a factor of three, and everything's fine. For other workloads, they need to get all the resources we promised them. Sorry not to give an exact answer; it's a rather similar situation to "is this machine big enough to run my database server?" but with more variables. - Peter On Tue, 27 Aug 2019 at 17:20, Kaushal Shriyan <kaushalshriyan@gmail.com> wrote:
On Tue, Aug 27, 2019 at 1:05 PM Sven Schwedas <sven.schwedas@tao.at> wrote:
That depends on what you plan on doing with the host. IIRC live VM migrations use host CPU time, and depending on transport can use quite a bit of CPU (for encryption/compression). Same with storage, if you have a ZFS/btrfs/LVM2/RAID/encryption setup that requires a lot of CPU, that's also counted against host CPU time.
libvirtd itself doesn't need all that much resources for itself, that said. 2 cores and 2GB RAM should suffice as baseline? Plus whatever you need to meet above needs, if any apply.
Hi Sven,
So for host OS, 2 vCPU's with 2GB Memory is enough. Please comment. I have spawned 10 KVM based VM's with CentOS 7.6 on the bare metal server (Dell R630 Poweredge 1U with 32 cores vCPU's and 96 GB RAM.) Is there a way to find out the system resources utilization for all the 10 VM's?
Best Regards, _______________________________________________ libvirt-users mailing list libvirt-users@redhat.com https://www.redhat.com/mailman/listinfo/libvirt-users