promox pools

[skraw]

Hello all,

I tried to find a way in proxmox to "bundle" vms together in a group which can then be addressed with functions like a single vm. Like I want to be able to migrate a pool from one node to another in a cluster. I would like to see it more like a group of vms rather than a "pool" of vms for administration or access purposes.
I found the "pveum pool" stuff, but did not see any hints if or how such a pool can be migrated together or shutdown together or the like.
How about that?

Regards

 

[bbgeek17]

Thats not a functionality that exists today. Migration in CLI or API only accepts single VM as target. Since GUI uses API its also limited to that.
You can issue many concurrent migrations (ie using GNU parallel), but you'd still have to address one VM per request.

You can make use of the tags to group the VMs, or perhaps even Pool, but you need two steps: first to gather list of VMs, then to migrate them.

FYI: https://kb.blockbridge.com/technote/proxmox-concurrent-vm-migration/index.html

PS note that there is a "Bulk Migrate" wizard available off-the-right click on the node. It has most if not all the functionality you are looking for. But it probably still issues multiple parallel requests, one per VM. I am guessing its essentially a wrapper to make admin's life easier.

---

Blockbridge: Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[skraw]

Ok, just to make that clear: my major concern is _not_ to bulk migrate many vms in parallel. I only do not want to click (or type) a hundred vm ids for migrating from one node to another. _AND_ - another point worth mentioning - I want to have them all together after this initial migration to be able to migrate them back to the original node. I really cannot understand why nobody ever thought about that clearly, because this is something that has to be done before almost every relevant node upgrade. So every proxmox user should have the need to do exactly that sooner or later.
If you migrate about 100 vms to another node already containing a significant number of vms you have a high time finding them again for back-migration.
How do you do that?

PS: Honestly I would want to have serial migration and not concurrent, be it only for safety reasons...
PS2: This function is of course available in VMWare (which I do not like for its flakey history)

 

[JasonJoel]

Assuming you are using HA on that size cluster, shouldn't HA should just "take care" of this if you put the host into maintenance mode?

From the docs ( https://pve.proxmox.com/pve-docs/chapter-ha-manager.html#_maintenance_mode ):

"The target nodes for these migrations are selected from the other currently available nodes, and determined by the HA group configuration and the configured cluster resource scheduler (CRS) mode. During each migration, the original node will be recorded in the HA managers' state, so that the service can be moved back again automatically once the maintenance mode is disabled and the node is back online."

 

[UdoB]

I only do not want to click (or type) a hundred vm ids for migrating from

The current "Bulk Actions --> Bulk Migrate"-Dialog allows exactly this: you can select all VMs which are member of a pool and/or you can select all VMs with a specific tag.

Best regards
Edit: probably Web-Gui only...

 

[skraw]

Assuming you are using HA on that size cluster, shouldn't HA should just "take care" of this if you put the host into maintenance mode?

From the docs ( https://pve.proxmox.com/pve-docs/chapter-ha-manager.html#_maintenance_mode ):

"The target nodes for these migrations are selected from the other currently available nodes, and determined by the HA group configuration and the configured cluster resource scheduler (CRS) mode. During each migration, the original node will be recorded in the HA managers' state, so that the service can be moved back again automatically once the maintenance mode is disabled and the node is back online."

I will not mess around with HA any more, did that with a cluster of 5 nodes and it always ended up the wrong way round.
This time we have a cluster of two nodes and I doubt this will make HA any better (doubting that it will work at all as I seem to remember you need at least three nodes for HA).

 

[skraw]

The current "Bulk Actions --> Bulk Migrate"-Dialog allows exactly this: you can select all VMs which are member of a pool and/or you can select all VMs with a specific tag.

Best regards
Edit: probably Web-Gui only...

Thank you for pointing to tags. After looking at some videos I tried that and it does as you say. It is not really nice looking, I personally would prefer a tree view where you can see the node and then a group and below it the vms. So the tag view is kind of second best.
Nevertheless I found out that tag attachment to a vm has serious problems. If you do not select the case sensitivity and mistype a tag first time during attachment it will always come out wrong, no matter how you type it. This is weird at least.

 

[JasonJoel]

Well, you were given multiple ways to do what you asked for - with today's functionality.

If you would prefer to put in a feature request and hope it gets done in the future, that is your choice of course.

Good luck!

 

[skraw]

I am no guy for a feature request of this kind. I rather prefer writing stuff myself, which I do since the 1980's.
The things I see missing in this issue are really marginal in terms of additional code. I believe they are not the coders' problem but rather a deficiency of the people defining the product. They have not looked at competitor solutions being a lot more straight forward and easy-to-use.

I personally find it more interesting to pinpoint why a kvm64 cpu on AMD is not compatible to a kvm64 cpu on Intel, obviously a qemu issue.
Lets end this thread here. Thank you.

 

[tom]

... Like I want to be able to migrate a pool from one node to another in a cluster.

Just use "Bulk migrate", seems the perfect fit for this.

 

[skraw]

Hi Tom,
Thanks for answering. Maybe I should have explained a bit more around the basic problem. It's not just aimed at the pure function itself, but rather how you can group/pool/tag some vms together to perform some function on them afterwards.
I did the tagging and it has really some peculiarities. Just try for yourself to create a tag, lets call it "Test'. Then try to add it and mistype it as "test". You get a tag with a different color. Now delete it and try to retype it as "Test". You will find that the tag is not added but instead the misspelled ''test" reappears. You have to enable case sensitivity in the options. Only then you can get to your predefined tag.
The whole thing is overly complex in usage. It would be a lot better to be able to create a new pool inside the tree view on the left and then simply drag the vms into this pool. Same applies for tags. Sure it somehow works currently, but compared to other existing implementations of such features it is far away from being user-friendly.

 

[tom]

If you have suggestions for improvements, please file them on https://bugzilla.proxmox.com - only request filed will be taken into consideration.

 

[skraw]

Sorry, bugfixes are solely your job, not mine. And I can see quite some necessary in your GUI.

 

[tom]

Sorry, bugfixes are solely your job, not mine. And I can see quite some necessary in your GUI.

Really? You request features but you reject to file them so that devs can take a look on it.

=> Your posts are just a big waste of time.

 

[skraw]

For me obviously, as you even reject to read them in detail...

 

[tom]

For me obviously, as you even reject to read them in detail...

If you want features or bugs fixes, you need to talk to the developers, best by filing your request into https://bugzilla.proxmox.com

I am a moderator here, and you are on a good way to get blocked as you ignore basic rules in our community.

 

[UdoB]

Sorry, bugfixes are solely your job, not mine.

That's probably true ;-)

The point of the last few posts was: This forum is mainly used by community members (though actively supported by several staff members, thanks!) so to make the actual developers aware of your idea you may put it on https://bugzilla.proxmox.com/ as a "Feature request". (Not as an actual bug.)

Best regards