We have about 15 PM servers running at 3 data center locations. Most are PM 1.8 and since some of the data centers have colocated servers and are remote (ie. 1,500 miles away), we don't get to the site location regularly. We have had no major issues with PM VE. Most of our guests are OpenVZ servers, and they run extremely fast and with low resource overhead.
We are now adding some Windows KVM servers to the fold, and they have issues with VirtIO drivers on PM 1.8. I believe the issues are related to the kernel versions, so we are using PM 2.2 for these servers.
This ends up in a monitoring challenge. We could cluster all of our PM 1.8 servers together and have been able to use a unified server monitoring solution with a browser for this by doing some modifications to the PM 1.8 web interface. However with PM 2.2, this is going to be more of a challenge.
I'm wondering how others who have multiple PM VE servers are doing this? How do you handle multi-locational hosting where hypervisor upgrades may not be easy or practical? Do you upgraded every PM VE server everytime a new PM VE is released? Do you wait for a while, and end up in the same situation that we are - different PM VE versions on different hardware boxes?
And how do you solve the need for a unified server monitoring solution? Something that can simply show the load and availability of guests on different PM VE server versions?
Thanks in advance for any feedback.
Myles
We are now adding some Windows KVM servers to the fold, and they have issues with VirtIO drivers on PM 1.8. I believe the issues are related to the kernel versions, so we are using PM 2.2 for these servers.
This ends up in a monitoring challenge. We could cluster all of our PM 1.8 servers together and have been able to use a unified server monitoring solution with a browser for this by doing some modifications to the PM 1.8 web interface. However with PM 2.2, this is going to be more of a challenge.
I'm wondering how others who have multiple PM VE servers are doing this? How do you handle multi-locational hosting where hypervisor upgrades may not be easy or practical? Do you upgraded every PM VE server everytime a new PM VE is released? Do you wait for a while, and end up in the same situation that we are - different PM VE versions on different hardware boxes?
And how do you solve the need for a unified server monitoring solution? Something that can simply show the load and availability of guests on different PM VE server versions?
Thanks in advance for any feedback.
Myles