I'm looking for a nice, elegant solution to (what I think) is a typical data center problem. I'm looking to setup something like the following:
2 OpenFiler boxes with 8 x SATA drives each on 5805z controllers in RAID config (RAID-6? or large RAID-10 probably). These will be the /var/lib/vz directories mounted via NFS (most likely) to the front-ends below. I'd like to run DRBD to mirror the two and implement some sort of HA so that at least one of the two is always up.
4+ "front-ends" that will run the machines. The machines will be split up among the currently running machines with the intention that if any one of the machines goes down for any reason, I simple have to copy the vz.conf file over to another machine and vzctl start it (since all machines will be NFS mounted).
Does this work? Is this the best way to handle this situation? I am running about 50 machines currently. With gigabit ethernet be fast enough to run these machines across the net? How easy is it to bond multiple gig-e cards together?
2 OpenFiler boxes with 8 x SATA drives each on 5805z controllers in RAID config (RAID-6? or large RAID-10 probably). These will be the /var/lib/vz directories mounted via NFS (most likely) to the front-ends below. I'd like to run DRBD to mirror the two and implement some sort of HA so that at least one of the two is always up.
4+ "front-ends" that will run the machines. The machines will be split up among the currently running machines with the intention that if any one of the machines goes down for any reason, I simple have to copy the vz.conf file over to another machine and vzctl start it (since all machines will be NFS mounted).
Does this work? Is this the best way to handle this situation? I am running about 50 machines currently. With gigabit ethernet be fast enough to run these machines across the net? How easy is it to bond multiple gig-e cards together?