Please forgive the dumb question(s).
Hypothetical setup:
3 servers, set to cluster, and in quorum.
Each have their own LVM volumes where volumes are mounted with EXT4 filesystems so directories and disks images are readable/viewable as files on the file system, i.e. each server has a mounted filesystem call "/vms" but may have different sized volumes but enough space for a set of VMs (more that 500GB), and that directory has been marked as "shared" in the web interface and each server has "/vms" on each system.
For instance, VM 101's disk image exists on all systems along with the typical directories of dump, images, private and template, yet is running on 1 node at a time.
If all disk images reside on each of the servers in the quorum, what issues might arise with live migrations?
Do the disk images sync up while migrating?
Are there any guest OS considerations to take into account?
I realize it's probably not "efficient", looking more for reliability with limited resources. I have searched but haven't found anything like this answered. Most things I find are for setting up DBRD, NFS and have some centralized storage, which I can't say will be an option.
That being said. I have a virtual cluster setup of this type for testing before going live with a hardware setup and it "seems" to work. I'm looking for some advice into how to make this work, or if it's a fools errand. Apologies if it comes off as hand holding.
Thanks for taking the time to read this, and apologies if this is not something, "obvious" that I'm not getting.
Hypothetical setup:
3 servers, set to cluster, and in quorum.
Each have their own LVM volumes where volumes are mounted with EXT4 filesystems so directories and disks images are readable/viewable as files on the file system, i.e. each server has a mounted filesystem call "/vms" but may have different sized volumes but enough space for a set of VMs (more that 500GB), and that directory has been marked as "shared" in the web interface and each server has "/vms" on each system.
For instance, VM 101's disk image exists on all systems along with the typical directories of dump, images, private and template, yet is running on 1 node at a time.
If all disk images reside on each of the servers in the quorum, what issues might arise with live migrations?
Do the disk images sync up while migrating?
Are there any guest OS considerations to take into account?
I realize it's probably not "efficient", looking more for reliability with limited resources. I have searched but haven't found anything like this answered. Most things I find are for setting up DBRD, NFS and have some centralized storage, which I can't say will be an option.
That being said. I have a virtual cluster setup of this type for testing before going live with a hardware setup and it "seems" to work. I'm looking for some advice into how to make this work, or if it's a fools errand. Apologies if it comes off as hand holding.
Thanks for taking the time to read this, and apologies if this is not something, "obvious" that I'm not getting.
