Hi there,
our about 50 server VMs are looking for a new home!
The current usage of resources goes like this: 192 vCPU cores (real usage on the two nodes is about 15-20%), 377 GB vRAM, 12 TB vDisk (assigned), 8,3 TB (used).
The load profile is a mix of everything: databases, mail, file and a lot of smaller stuff. Guest operating systems are Windows (Server 2012 R2) and Linux (Ubuntu Server, Oracle Linux)
So I guess the standard idea is building a three node PVE+Ceph cluster. That would be fine, but has one big disadvantage: Moving from the current 2x2 CPU environment to a 3x2 CPU environment means paying much more money to Oracle and Microsoft.
This is why I'm looking for alternatives.
What do you think about this?
We build two big PVE machines which are not clustered. All VMs live on one of them all the time, and the second machine acts as a cold standby which continuously gets storage sync and everything else what is needed to bring the VMs into business in case of emergency.
My first question here is: If we sync the storage "incrementally" every minute, the amount of data transfered would be little. But I'm worried about the process finding out what data must be transfered related to time and storage/CPU usage.
Thanks a lot for your support and many greets
Stephan
our about 50 server VMs are looking for a new home!
The current usage of resources goes like this: 192 vCPU cores (real usage on the two nodes is about 15-20%), 377 GB vRAM, 12 TB vDisk (assigned), 8,3 TB (used).
The load profile is a mix of everything: databases, mail, file and a lot of smaller stuff. Guest operating systems are Windows (Server 2012 R2) and Linux (Ubuntu Server, Oracle Linux)
So I guess the standard idea is building a three node PVE+Ceph cluster. That would be fine, but has one big disadvantage: Moving from the current 2x2 CPU environment to a 3x2 CPU environment means paying much more money to Oracle and Microsoft.
This is why I'm looking for alternatives.
What do you think about this?
We build two big PVE machines which are not clustered. All VMs live on one of them all the time, and the second machine acts as a cold standby which continuously gets storage sync and everything else what is needed to bring the VMs into business in case of emergency.
My first question here is: If we sync the storage "incrementally" every minute, the amount of data transfered would be little. But I'm worried about the process finding out what data must be transfered related to time and storage/CPU usage.
Thanks a lot for your support and many greets
Stephan
So I will mark this thread as "solved" and open another with all my ideas and questions concerning our future environment.