Does migration even work ? Storage 'xxx' not available, but can't duplicate !

[mike1033]

Does migration on Proxmox works ?

I've found online that people advising to backup => copy => restore...

Does it means that Replication is broken and in a future will be added ?

Currently requirements cannot be logically met as it requires the same storage name on different machine. This of course won't happen as you cannot create storage with the same name in a cluster.

Extremely annoying to find it out after moving from ESXI, as this was ONLY reason for us to move on Proxmox.

Am I something missing ?

Server 007 has storage hdd_b.

Now I want to move VM 100 from 007 to 004. However I cannot create storage with the same name 'hdd_b' as I am getting info about duplicate...

Unfortunately cannot backup as for some reason Proxmox on 007 stolen all hard drive (300GB) and local is only 9GB (only on 007, but this is a story for another topic).


Can anyone explain to me, how replication works ?

I've tried to clone it only to see that it requires same target storage (???) & that 004 is not allowed for this action.

 

[dietmar]

Does migration on Proxmox works ?

yes, that works perfectly.

I've found online that people advising to backup => copy => restore...

That also works.

Does it means that Replication is broken and in a future will be added ?

nothing is broken.

Currently requirements cannot be logically met as it requires the same storage name on different machine. This of course won't happen as you cannot create storage with the same name in a cluster.

What's the problem. We all use it that way and it works perfectly.

Extremely annoying to find it out after moving from ESXI, as this was ONLY reason for us to move on Proxmox.

What exactly is annoying?

Am I something missing ?

Already read the documentation? See https://pve.proxmox.com/pve-docs/

Server 007 has storage hdd_b.

Now I want to move VM 100 from 007 to 004. However I cannot create storage with the same name 'hdd_b' as I am getting info about duplicate...

You use the same storage name on all cluster nodes. Just mount the disk to the same file path (or use the same LVM names)

 

[mike1033]

yes, that works perfectly.

Hmmm

Already read the documentation? See https://pve.proxmox.com/pve-docs/

You meant this ?
https://pve.proxmox.com/pve-docs/pve-admin-guide.html#qm_copy_and_clone

It is as useless as it gets. Documentation on a proxmox is very bad, a lot of broken/empty pages and very vague more a description than actual documentation (though I am software engineer and I am used to perhaps a different quality).

You use the same storage name on all cluster nodes. Just mount the disk to the same file path (or use the same LVM names)

Doesn't work.

Server004 => has hdd_b storage (created using vgcreate, as wasn't able to use GUI to create a new one after I've removed previous LVM, was giving me only no unused drives, yes was using "documentation" to recreate)
Server007 => has vm 100 that is using hdd_b lvm on Server007. Now when I try to migrate it it will see the lvm ID, not the name !!! And LVM ID cannot be the same ! ("crate storage failed: error with cfs lock 'file-storage_cfg': storage ID 'hdd_b' already defined (500)").

So the LVM name doesn't mean anything for migration, as it use only ID, those are completely different properties (name, id)

Could you explain how this work in a real world then ?

Looks like a huge issue with the feature as why I should create LVM named exactly same on 10 or 20 servers ???? LVMs are named after the actual bay and I cannot have 25 drives named HDD just in case I would want to migrate/clone VM to one of them.

Seems like it was designed for homelabs and not actual commercial applications as have 10x server with at least 16 drives and naming each drive in the same way so one of the most important features would work, seems like simply and plainly stupid idea. Why would I be forced to having 160 LVMs named in the same way ? This adds extra layer during architecture planning (hardware & software should be de-coupled, I shouldn't be thinking about LVM names just to move freely VMs between environments.

Simple example

Our platform is using > 15 applications, each application can have up to 5-10 instances (for spreading the load), plus we have 5 environments (2 live and 3 testing).

Do you see how silly it is to have such a constrains on your system ? Did you ever talked with real stakeholders or just let anyone to create new constrains on certain feature and just let's figure it out later.

It is failed project because of that and not one company will use it (at least one with > 1 instance of the application per environment).

It seems like a nightmare as we keep adding servers and having this constrain in a place would be a huge headache.

I was hoping that we could replace ESXI, but you are nowhere close even to free ESXI.

Total amateurish piece of software, which is shame as you did huge amount of work, which is wasted by lack of proper DevOps and proper plan for development.

 

[tom]

Seems you got a bad start with Proxmox VE. But instead of telling we are amateurs or doing bad software, try to find out why it does not work in your setup. All what not working on your setup is working on ten thousands places elsewhere ...

 

[spirit]

I'm running 4000 vm in production with this "amateurish piece of software,", 200TB ssd storage with ceph.

so maybe RTFM before complain


If you only use local storage (so, not a production storage because you can't failover vm),
you can migration vm + migrate local storage at the same time, but command line only:

qm migrate <vmid> <targetnode>--with-local-disks --targetstorage <optionnal_diferent_local_storagename_on_targetnode>


also:

#man qm

 

[brosky]

I've bumped in the same situation and let me explain it with easier words so that everybody can understand.

When you have a cluster you will need to have some resources "shared" , in this case the storage paths.

Look at the /etc/pve/storage.cfg - there you will have the storage paths defined for the entire cluster - and this means that the same paths must exist on each node.

In my case, i have:
node1

/ssd for a raid1 SSD's (2x500Gb)
/hdd for a raid1 HDD's (2x8T)

On node 2, even if I have a different hardware configuration, i've mounted the arrays on the same paths to have consistency:

/ssd for a raid6 (5x500G)
/hdd for a raid10 (4x8T)

So, even if you have different hardware architectures (software or hardware raid, plain disks, etc) you will need to have the same mounted paths.

You'll see that the "Datacenter->Storage" menu does not give you a per-node configuration but per-cluster.

Hope this helps.

 

[spirit]

I've bumped in the same situation and let me explain it with easier words so that everybody can understand.

When you have a cluster you will need to have some resources "shared" , in this case the storage paths.

Look at the /etc/pve/storage.cfg - there you will have the storage paths defined for the entire cluster - and this means that the same paths must exist on each node.

In my case, i have:
node1

/ssd for a raid1 SSD's (2x500Gb)
/hdd for a raid1 HDD's (2x8T)

On node 2, even if I have a different hardware configuration, i've mounted the arrays on the same paths to have consistency:

/ssd for a raid6 (5x500G)
/hdd for a raid10 (4x8T)

So, even if you have different hardware architectures (software or hardware raid, plain disks, etc) you will need to have the same mounted paths.

You'll see that the "Datacenter->Storage" menu does not give you a per-node configuration but per-cluster.

Hope this helps.

What is nour need exactly ?

my personal opinion is that local storage should be define at node level, not datacenter level, because it's confusing. (but it's not implemented like this in proxmox)

But, anyway, 1 storage defined in storage.cfg, if you need live migration,you really need 1 real shared storage (nfs,rbd, cifs, shared lvm,..), so only 1 path can exist.


So, for special local case:

you can have 2 zfs local storage with async replication. (but no live migration).
you can have 2 different storagename/zfspool on each storage, proxmox manage that for replication, and will change vm config when HA if enabled.



you can also do livemigration+storagemigration in proxmox6, it's working with 2 differents local storagename too.