Backup to anoher server and NFS mount

[driftux]

Hello, I'm very new in Proxmox and I have very important question to ask, because I'm very carefully planning the backup solution. I would like to ask is it possible to configure Proxmox and it's Backup server according to the situation below:

1) Server A with Proxmox installed makes backup to another Proxmox server B (separate disk array for backups) over the local network.

2) Server A damages and it goes offline.

3) Server C with Proxmox is turn on and I manually mount Server B directory with VM backups (VM disks) and this combination of server C and B starts serving like it will be Server A. I mean, as the example, I need in a very short time attach Proxmox to NAS and use VM backups (images) from there. Can I fulfill such scenario with Proxmox Backup Server?

My purpose is that two Proxmox servers A and B is always on and have some different virtual machines running. Both servers are using 90% of resources and they are doing backup to each other. When the situation comes that one of the servers is damaged and can't be repaired quickly, the server C is started. Server C mounts the backup (VM disks) and and work like a quick disaster recovery solution.

And the second question about such case mentioned above performance. What do you think, will NFS connection between Server B (VM storage) and Server C (VM host) can handle the load of 1000 website and 5000 mailboxes on KVM machine through 1Gbps network? Won't be troubles in such scenario? I need that several week it will serve for the customers very flawlessly while server A be repaired.

 

[oversite]

Hello, I'm very new in Proxmox and I have very important question to ask, because I'm very carefully planning the backup solution. I would like to ask is it possible to configure Proxmox and it's Backup server according to the situation below:

1) Server A with Proxmox installed makes backup to another Proxmox server B (separate disk array for backups) over the local network.

2) Server A damages and it goes offline.

3) Server C with Proxmox is turn on and I manually mount Server B directory with VM backups (VM disks) and this combination of server C and B starts serving like it will be Server A. I mean, as the example, I need in a very short time attach Proxmox to NAS and use VM backups (images) from there. Can I fulfill such scenario with Proxmox Backup Server?

My purpose is that two Proxmox servers A and B is always on and have some different virtual machines running. Both servers are using 90% of resources and they are doing backup to each other. When the situation comes that one of the servers is damaged and can't be repaired quickly, the server C is started. Server C mounts the backup (VM disks) and and work like a quick disaster recovery solution.

And the second question about such case mentioned above performance. What do you think, will NFS connection between Server B (VM storage) and Server C (VM host) can handle the load of 1000 website and 5000 mailboxes on KVM machine through 1Gbps network? Won't be troubles in such scenario? I need that several week it will serve for the customers very flawlessly while server A be repaired.

Answer to first question, perhaps running both proxmox and pbs on server c, you backup both a and b to c, if either a or b fails, restore on c from c datastore and run VMs on c.

Second question, you mean you have the VMs disks on NFS storage on 1Gbit? It will certainly work but it will be a bit slow for the VMs. Probably not painfully slow.

 

[driftux]

Yes, I thought about holding every server backup on C, but there is a huge problem, that when I will have 10-20 server, my C server should have enormous count of HHD drives and space. So I think about the solution, that every server would be like a NAS for his server brother. The only problem is here that if both brothers will go offline. So can I go this way?

Do I understand correctly, that I only need to push the restore button on C server and selected the latest possible snapshot, and I will be ready to serve my clients in a minutes? I mean I don't want to pickup 6TB machine and wait several hours till it restores. I want to restore the snapshot as the Veeam software does on ESXi.

 

[oversite]

Yes, I thought about holding every server backup on C, but there is a huge problem, that when I will have 10-20 server, my C server should have enormous count of HHD drives and space. So I think about the solution, that every server would be like a NAS for his server brother. The only problem is here that if both brothers will go offline. So can I go this way?

Do I understand correctly, that I only need to push the restore button on C server and selected the latest possible snapshot, and I will be ready to serve my clients in a minutes? I mean I don't want to pickup 6TB machine and wait several hours till it restores. I want to restore the snapshot as the Veeam software does on ESXi.

I don't know Veeam and ESXi. A restore of a 6TB VM will take hours depending on your hardware.

 

[driftux]

Let's simplify the scenario. Can I have Proxmox server with active VM, and make replication of that VM, to another Proxmox server. We could say that I will have active VM on one server, and hot stand-by VM copy on another server. As understood Proxmox backup server initialy will copy all VM data to another server, and after this action it only will send snapshots of active machine by intervals. In case of failure with the main server, I connect to another server, when select the last good snapshot and just start the hot stand-by machine, because it's a clone with only several minutes lost of data. So in this scenario, I can start 6TB machine in minutes. Do I understand correctly?

If I'm right in this case, I can have server X, with active VMs and hot stand-by VMs from server Y.
Server Y will have it's own active VMs, and have hot stand-by VMs from server X.

If one of my servers will go offline and I will need to repair it hardware, I only need to start stand-by VMs on it's brother.
Veeam backup is working the same as I wrote in this scenario.

Ok, now please imagine that both servers running 90% of all power. So in case of failure, I can't star the machine on the Brother server, because it won't handle the load. So I think to have one more phisical server which will be turned on disaster, and mounts through NFS the VM from the Brother server.

Later I will expand my servers adding every time by 2 servers. And I always have one server which is shut down and only will be powered on to mount the VMs.

Please tell me is this scenario is possible?

 

[oversite]

Let's simplify the scenario. Can I have Proxmox server with active VM, and make replication of that VM, to another Proxmox server. We could say that I will have active VM on one server, and hot stand-by VM copy on another server. As understood Proxmox backup server initialy will copy all VM data to another server, and after this action it only will send snapshots of active machine by intervals. In case of failure with the main server, I connect to another server, when select the last good snapshot and just start the hot stand-by machine, because it's a clone with only several minutes lost of data. So in this scenario, I can start 6TB machine in minutes. Do I understand correctly?

If I'm right in this case, I can have server X, with active VMs and hot stand-by VMs from server Y.
Server Y will have it's own active VMs, and have hot stand-by VMs from server X.

If one of my servers will go offline and I will need to repair it hardware, I only need to start stand-by VMs on it's brother.
Veeam backup is working the same as I wrote in this scenario.

Ok, now please imagine that both servers running 90% of all power. So in case of failure, I can't star the machine on the Brother server, because it won't handle the load. So I think to have one more phisical server which will be turned on disaster, and mounts through NFS the VM from the Brother server.

Later I will expand my servers adding every time by 2 servers. And I always have one server which is shut down and only will be powered on to mount the VMs.

Please tell me is this scenario is possible?

No, you cannot use proxmox backup server for this. Your scenario works with a shared vm storage, like nfs or ceph, but as you have understood the nfs storage also need another replica to not be single point of failure. This can be achieved but not using PBS. The snapshot-based chunks on PBS cannot be mounted as the VMs virtual disks, they have to be restored. Perhaps pve-zsync works for you, or you could look at ceph storage.

 

[driftux]

I thought that I can do active machine backup once in order to have a cold (not running) machine. Then just sent snapshots all the time from active machine for example every 15minutes to the cold machine. In active machine failure, I could restore my cold machine and have 15minutes old data.

I thought about scenario where all machine backups will reside on huge hdd volume and a powerfull server. So in case of failure this huge NAS will become a server when some VM will be restored. But more effective way is to have cold machines backups on various servers just a separate HHD space. In case of failure, this backup space would be manually mounted throught NFS from another server, which is power off all the time, and would turned on only in case of failure. I just need to mount my cold machine disk and power on VM through NFS.

I don't get it... Are you sure that both scenarios is impossible with Proxmox Backup Server?

 

[oversite]

I thought that I can do active machine backup once in order to have a cold (not running) machine. Then just sent snapshots all the time from active machine for example every 15minutes to the cold machine. In active machine failure, I could restore my cold machine and have 15minutes old data.

I thought about scenario where all machine backups will reside on huge hdd volume and a powerfull server. So in case of failure this huge NAS will become a server when some VM will be restored. But more effective way is to have cold machines backups on various servers just a separate HHD space. In case of failure, this backup space would be manually mounted throught NFS from another server, which is power off all the time, and would turned on only in case of failure. I just need to mount my cold machine disk and power on VM through NFS.

I don't get it... Are you sure that both scenarios is impossible with Proxmox Backup Server?

You are confusing me, i don't understand what you are asking when you say cold machine. Also, are you talking about containers or KVM VMs?
The proxmox backup server does not have the VMs KVM disks, it has the data stored in chunks, they cannot be mounted to run the VM, they need to be restored somewhere.
Obviously you can do backups with snapshots every 15 minutes and restore the VM on a proxmox host to run it.

 

[driftux]

My main aim is to have a two clones of VM. One is active and another is passive (cold). Active VM always send snapshots to the passive VM, and in case active machine of failure I just turn on the passive machine, which will become active with minimal data loss. I thought that this scenario is possible with PBS. Actually Veeam clone program can do this way. You can set Veeam to send replicas to another Vmware machine and if you need you can turn of the primary machine, and activate the clone without waiting restore process. So the same I thought is possible with PBS.

Ok, I thought that PBS works with KVM also, if not, then lets' talk about containers only.
I will ask in different way. For example I has 6TB container on Proxmox server. I do a single full backup or clone of container to another server with 8TB storage (6TB for clone container and ~2TB for snaphshots). I will never do a full backup no more, and only will send snapshots to another server from this moment. Because I can't send all the time 6TB over my LAN network.

So please tell me, can I just activate the clone container with some manual work and have system running in several minutes?
If not, the PBS only support backup feature, and I need in total 16TB of storage in my backup server, that to store container backup+snapshots(6+2TB) and 8TB empty space, and then only to start restore process. So in the conclusion of the restore, I will have 8TB backup+snapshots and 8TB occupied with the restored machine. So this means that PBS is only for backup process but not for the cloning scenario.
Am I correct?

 

[driftux]

I just open documentation FAQ:

5.5 Can I copy or synchronize my datastore to another location?
Proxmox Backup Server allows you to copy or synchroize datastores to other locations, through the
use of Remotes and Sync Jobs. Remote is the term given to a separate server, which has a datastore
that can be synced to a local store. A Sync Job is the process which is used to pull the contents of a
datastore from a Remote to a local datastore.

So this very near what I want, but I asked about container cloning, now I read that I can clone not the machines, but the datastores. So in case of failure, I can connect to another server with cloned datastore and create virtual machine there and just add a disk to VM. Maybe I don't even add a disk or just press the button to start container. So this means that in several minutes I can restore 6TB container. Or am I still missing something here?

 

[driftux]

Anyone?

 

[oversite]

I just open documentation FAQ:

5.5 Can I copy or synchronize my datastore to another location?
Proxmox Backup Server allows you to copy or synchroize datastores to other locations, through the
use of Remotes and Sync Jobs. Remote is the term given to a separate server, which has a datastore
that can be synced to a local store. A Sync Job is the process which is used to pull the contents of a
datastore from a Remote to a local datastore.

So this very near what I want, but I asked about container cloning, now I read that I can clone not the machines, but the datastores. So in case of failure, I can connect to another server with cloned datastore and create virtual machine there and just add a disk to VM. Maybe I don't even add a disk or just press the button to start container. So this means that in several minutes I can restore 6TB container. Or am I still missing something here?

Driftux, yes you are still missing something-
I really suggest you install PBS and play with it, that will give you more answers than this forum. I have tried to tell you some key points, but i am afraid you don't really listen if you don't like the answers.
However, here are my two cents:
PBS is a fantastic piece of software, but it's not a fail over or high availability stack. You cannot do what you want with PBS. If you want a hot clone, don't use PBS for that. Use PBS for real backups that can be restored if disaster or finger trouble. You can mount them if you need a few files instead of a full restore, but you cannot use the chunks to run your VMs. If you don't restore but map them, possible with bad performance you could in theory run something, but not as the primary VMs. Proxmox staff can correct me if i am wrong.

For your hot/cold VMs, I told you before, dunno why you showed no interest in it, you might need the VMs on shared storage for easy fail over, like CEPH or even NFS. If you don't want to go that road, look at zfs / pve-zsync / replication in PVE GUI.

 

[christophe]

My main aim is to have a two clones of VM. One is active and another is passive (cold). Active VM always send snapshots to the passive VM, and in case active machine of failure I just turn on the passive machine, which will become active with minimal data loss. I thought that this scenario is possible with PBS. Actually Veeam clone program can do this way. You can set Veeam to send replicas to another Vmware machine and if you need you can turn of the primary machine, and activate the clone without waiting restore process. So the same I thought is possible with PBS.

Ok, I thought that PBS works with KVM also, if not, then lets' talk about containers only.
I will ask in different way. For example I has 6TB container on Proxmox server. I do a single full backup or clone of container to another server with 8TB storage (6TB for clone container and ~2TB for snaphshots). I will never do a full backup no more, and only will send snapshots to another server from this moment. Because I can't send all the time 6TB over my LAN network.

So please tell me, can I just activate the clone container with some manual work and have system running in several minutes?
If not, the PBS only support backup feature, and I need in total 16TB of storage in my backup server, that to store container backup+snapshots(6+2TB) and 8TB empty space, and then only to start restore process. So in the conclusion of the restore, I will have 8TB backup+snapshots and 8TB occupied with the restored machine. So this means that PBS is only for backup process but not for the cloning scenario.
Am I correct?

Hi,

you don’t need pbs to do that if i understand. You need a cluster, and a shared storage.
you could setup HA, and a node failure will trigger vm start on another note.

Christophe.