Hello,
hope my questions won't be too silly. I've tried cluster config of two Proxmox PCs with DRBD (though my DRBD performance was too poor, I believe it just question of tuning it up) and I've found ability to live migrate VMs from one host to another is something really nice.
But beside that "magic" feature, I've found cluster still lack of some badly needed things, and that was the reason for me to refuse to use it. But maybe you tell me I'm wrong and there is a way out, so I can somehow use cluster (and that's what I'd like to)?
Here are the things I've found about cluster:
1. Even though I use cluster (1 host is master, another is slave) and both hosts are use the same disks (via DRBD the storage is shared), I find no easy way to control cluster when master is dead. I used to put some like 3 VMs on 1st host (master), and another 3 VMs on 2nd host (slave), and if I loose master, I see no way to migrate VMs from it to slave (and even simple restart these VMs on slave) - slave refuse my commands saying I need to execute my commands on master.
What I'd like to see is some HA-daemon to control all hosts in cluster so in case of any host dying VMs from it to move (live or not) to other hosts.
2. I find no way to "freeze" VM disk state at the moment so I can "rewind" to it later. At least GUI won't give me that, and I find no recommended way to do that either; perhaps I could use LVM's standard snapshots but I won't doing whole storage snapshot, just one machine; I used classic "VM disk over LVM over DRBD" scheme, and I just don't know if I can simple snapshot single "VM disk".
3. It is fuzzy for me if I can add LVM stogage as backup one? Say, if I make DRBD backup disk, I can't use it as backup destination drive. Now the backup is connected to master only, and if master goes offline, I can't access backups.
Thank you in advance for your patience and hope I'll fine my answers!
hope my questions won't be too silly. I've tried cluster config of two Proxmox PCs with DRBD (though my DRBD performance was too poor, I believe it just question of tuning it up) and I've found ability to live migrate VMs from one host to another is something really nice.
But beside that "magic" feature, I've found cluster still lack of some badly needed things, and that was the reason for me to refuse to use it. But maybe you tell me I'm wrong and there is a way out, so I can somehow use cluster (and that's what I'd like to)?
Here are the things I've found about cluster:
1. Even though I use cluster (1 host is master, another is slave) and both hosts are use the same disks (via DRBD the storage is shared), I find no easy way to control cluster when master is dead. I used to put some like 3 VMs on 1st host (master), and another 3 VMs on 2nd host (slave), and if I loose master, I see no way to migrate VMs from it to slave (and even simple restart these VMs on slave) - slave refuse my commands saying I need to execute my commands on master.
What I'd like to see is some HA-daemon to control all hosts in cluster so in case of any host dying VMs from it to move (live or not) to other hosts.
2. I find no way to "freeze" VM disk state at the moment so I can "rewind" to it later. At least GUI won't give me that, and I find no recommended way to do that either; perhaps I could use LVM's standard snapshots but I won't doing whole storage snapshot, just one machine; I used classic "VM disk over LVM over DRBD" scheme, and I just don't know if I can simple snapshot single "VM disk".
3. It is fuzzy for me if I can add LVM stogage as backup one? Say, if I make DRBD backup disk, I can't use it as backup destination drive. Now the backup is connected to master only, and if master goes offline, I can't access backups.
Thank you in advance for your patience and hope I'll fine my answers!