pve-zsync for whole servers, not just for specific VM, possible? && pve-zsync GUI.

[mailinglists]

Hi,

i'm replacing proxmox official backup system, which is a bit resource heavy with it's "copy whole VM" for each backup with pve-zsync which can sync only differences.

All PM nodes 5.* are in a cluster. One node is designated as backup node and will pull VM zvols from other nodes. Right now I have to specify each VM by it's ID and location. I tried syncing whole rpool/data, but did not do it recursively.

Because I migrate the VMs around the cluster quite a lot, manually changing pve-zsync after each move is too time consuming and prone to errors / forgetting.

Is there a way to just say to pve-zsync, look here is a node XY, here is another node, etc, .. just pull whathever zvols for VMs are there, to local disks?
This way, i would not have to deal with pve-zsync everytime I move a VM to another node or add a new VM.

On a semi related note, does anyone know when we can expect pve-zsync GUI. I saw it on the road map in 2016, but it's still is not here.

 

[wolfgang]

Hi,

On a semi related note, does anyone know when we can expect pve-zsync GUI. I saw it on the road map in 2016, but it's still is not here.

This feature called "asynchronous Storage Replication" and is included since 5.0

If you are in a cluster you should use Storage Replica which handles the offline migration automatically.

pve-zsync can only sync one VM/dataset with one job.

 

[mailinglists]

Thank you for your answer.
Too bad pve-zsync can do only one VM per job, but if below described sollution works, there should be no need for pve-zsync.
I decided to try pve-zsync, because when reading this forum, seem like it's more stable than pvesr.
I also thought "storage replication" in GUI uses pvesr and not pve-zsync and that they are two different systems (therefor each should have it's own GUI).

I guess I could use storage replication in GUI and then something like zfSnap on target to get daily and monthly backups.

I always move --online --with-local-disks. I know it's not stable but it works for me (sometimes I have to pause the heavy used VM beforehand). So taking that into account, I wonder how will "storage replication" handle me moving the VMs. Can we expect any problems in the following scenario described below?

For example let's say we use lots of nodes, which we replicate to node p27.
So I create replication of VM 100 from node p26 to p27.
I then live migrate via command line VM 100 to node p25.
Will the replication of VM 100 from new node p25 to node p27 still work?

 

[wolfgang]

Will the replication of VM 100 from new node p25 to node p27 still work?

No, because no snapshot will be on the target node.
Storage live migration is a qemu feature and makes a one to one copy of the internal raw image.
It does not know anything about snapshots.
pvesr and pve-zsync need snapshots do calculate the diff.

 

[mailinglists]

OK, so before moving/migrating a VM from a node, to another node I need to delete replication and create it again after VM was migrated?

 

[wolfgang]

OK, so before moving/migrating a VM from a node, to another node I need to delete replication and create it again after VM was migrated?

Yes, this is the way.

 

[guletz]

Hi,

In my own opinion, pve-zsync is more useful because it can be used even for 2 different proxmox cluster. Even for 2 non-cluster proxmox servers. For this reasons, I think that will be more useful that Proxmox guys to concentrate their time to pve-zsync even for the the fact that is very stable (rock solid if anyone ask me)

How it will be for exemple datacenter replication with something who is pve-zsync? It is possible to set custom zfs proprieties/tags(maybe I use the wrong term, because I do not have a pc around me). With this tags, it could be possible to do many things (recursive replication, retention period, reverse the replication if we move a VM, and more). With some kind of stuff, it will be possible (I guess) to do live migration for a VM from datacenter A to datacenter B, or even do a all datacenter migration with all VMs.
Even more multi data center replication, or multi VM replication. Maybe I want to replicate VM xxxx to 3 different datacenters using only some custom tags in zfs proprieties(like datacenter replication yes/no, VM datacenter replication yes/no, reverse replication yes/no).
Also let say we have a fast storage(ssd or nvme) and a slow storage in our datacenter. Sometime we need some burst for a short period. And we could set a custum zfs tag for our VM yyyy. We know that we need more power (resurces) only monday from 12.oo to 12.300! So we can move our VM from slow storage to fast storage during this time.

Please think at this stupid ideas!

Thx.

 

[guletz]

And the last thing ... why do you not run a question/survey like: hi all proxmox users, what you thing will be most important for you:

, 1, 2, 3, ....

Then the proxmox stuff can use their limited time for what your customers desire / need. Or mybe what is need to be improved (like backups, zfs, web interfaces, migration, and so on).
Another ideea is to create let say unofficial wiki with some tutorials about tasks that are very useful for many users (like how to use a usb disk for backups - my shame because I do no remeber the author ). This must be done (as many sites use this unofficial wiky) because is very time consuming to find some kind of solutions. For this reason I save this kind of posts on my PC, and I make some notes about this in plain text file.

Like I said, all are suggestions, and maybe some of them are wrong .

Thx. again