Multi Datacenter Management

[Falk R.]

Ignoring bugs for the moment, what will it not do, that vcenter does?

The main purpose for my customers will be to have many clusters and single nodes distributed across Europe or the whole world in one GUI. This provides a rudimentary health overview and makes it easy to centrally manage VMs in different clusters/nodes.
It is also very important to have centralized rights management to determine which admin can access which clusters / nodes or VMs.

Cluster management and configuration of the nodes is still done individually.

Central patch management might also be a good idea, but this will certainly not be in the first version.

 

[jlauro]

You have to understand part of the ethos of this community manifests itself in bringing up everything bad about the so-called competition that others are willing to pay for multiple $$$ for. That's about it.

NB If anything, it will be GUI wrapper around qm remote-migrate, I would not expect much more. You can't really have anything more with the corosync based model.

That, and a single plane of glass that consolidates all the clusters into one tree view is about all I expect.
Ideally can show all vms for all clusters at once with load and other summary, but as long as can drill down it's ok.

 

[esi_y]

That, and a single plane of glass that consolidates all the clusters into one tree view is about all I expect.
Ideally can show all vms for all clusters at once with load and other summary, but as long as can drill down it's ok.

Sometimes I wonder, how many users can actually appreciate the quorum based cluster setup. After all, one wants a control node (whatever you want to call it) and in that case that might have been the master instead of the clumsy quorum. If you lose it, fair enough, you are like without quorum, but it's rather simple setup and does not have special demands on networking latency, etc.

IOW Once there's single control plane for multiple clusters, that means also multiple nodes, that means also someone will wonder, why can't I run cluster in the same manner. Geographically scattered at that.

 

[jlauro]

Sometimes I wonder, how many users can actually appreciate the quorum based cluster setup. After all, one wants a control node (whatever you want to call it) and in that case that might have been the master instead of the clumsy quorum. If you lose it, fair enough, you are like without quorum, but it's rather simple setup and does not have special demands on networking latency, etc.

IOW Once there's single control plane for multiple clusters, that means also multiple nodes, that means also someone will wonder, why can't I run cluster in the same manner. Geographically scattered at that.

Having a single consolidated management pane is different that a cluster that needs low latency between nodes. I have about 7 locations with about 1000 vms spread from coast to cost in the US, and a single vcenter works well. Each location has 1 or more clusters.
Vcenter is nice in that it gives you an easy one place to manage everything, but if it's down you can still manage the hosts directly while vcenter is down. Cluster features such as HA failover will also continue to work while Vcenter is down as it's mostly a layer on top and although needed for doing the setup between hosts on a cluster, the cluster continues those functions without vcenter. Of course, how much cluster activities you can do is limited if vcenter is down, but the important thing is you still have the option do manage the individual hosts.

There are obvious reasons (well, maybe not to everyone) why you can't easily run a cluster the is geographically scattered. That said, there is no reason you shouldn't be able to move a template or a vm from one location/cluster to another, or deploy a vm from a template hosted in a another location (granted it will be slower, but should be possible from a single management pane).

 

[esi_y]

There are obvious reasons (well, maybe not to everyone) why you can't easily run a cluster the is geographically scattered.

Can you share those obvious to you?

 

[LnxBil]

Can you share those obvious to you?

One is most certainly latency.

 

[esi_y]

One is most certainly latency.

Of cluster operations? If you had to wait up to ~5 seconds for inter-region migration, would you not want to have it automated? There's a wait for self-fencing nowadays already, it's not STONITH.

 

[sierra_mike]

@t.lamprecht

Any updates? Would love to get this going. It will shave hours off my week once deployed.

 

[Proxar]

Important feature request: There should be an option to let a cluster manage another remote cluster, but not the other way. Maybe integrate it into the ACL system.
At work we manage many customers, each with their own Proxmox clusters or standalone nodes and it would be great to have one interface for all of them.

 

[XMarcR]

Important feature request: There should be an option to let a cluster manage another remote cluster, but not the other way. Maybe integrate it into the ACL system.
At work we manage many customers, each with their own Proxmox clusters or standalone nodes and it would be great to have one interface for all of them.

+1

 

[4f1sh3r]

I had the hope that the beta/preview would be in 8.3, which released today, but seems not the case, or is it hidden?

 

[sierra_mike]

I had the hope that the beta/preview would be in 8.3, which released today, but seems not the case, or is it hidden?

@t.lamprecht

Any updates? Anyone?

 

[t.lamprecht]

This won't just be a simple PVE feature, so it's not tied to a PVE release.

 

[sierra_mike]

This won't just be a simple PVE feature, so it's not tied to a PVE release.

As expected, but is there an idea of where it is in terms of testing or release?

 

[t.lamprecht]

As expected,

Then it would have been nice to avoid an all caps message in an unrelated thread...

but is there an idea of where it is in terms of testing or release?

None that will help you now, but we're working on it and once we got something in a presentable beta stage we will naturally release that ASAP.

 

[sierra_mike]

You are missing the point, and doing a poor job at delivering information.

I will just assume you dont actually care since you suggested I just do your job for you in the other thread.

Im outa here.

 

[t.lamprecht]

You are missing the point, and doing a poor job at delivering information.

Not sure what point you made, and it's hard to give some info out if there isn't any new one that would fit on a generic "any updates" request. And getting some snark back on an all-caps message on a rather unrelated thread isn't IMO really unexpected.

I will just assume you dont actually care since you suggested I just do your job for you in the other thread.

We already created the API that can be used to relatively easily automate a lot of tasks, leveraging that for the infrastructure you have to maintain, at least until a more integrated, battery included tooling comes along is certainly not my job. And while it might have been snarky in the context, it was not dishonest, we have many users that leverage the API and standard automation tooling to do some part of their work very efficiently, and not all of that will be replaced by a datacenter manager, at least not in the first version.

If you have more concrete questions (besides a specific release date, that info is not available yet) with context of why they are important to you then we can see if we actually give you some information for that, because as of now I'm not even sure what I'm allegedly not caring about.

 

[jlauro]

If you have more concrete questions (besides a specific release date, that info is not available yet) with context of why they are important to you then we can see if we actually give you some information for that, because as of now I'm not even sure what I'm allegedly not caring about.

What features are expected for the initial public beta release?
If different (ie: more features added during the beta process), what features are expected for the official release?
Will this be included as part of PVE subscriptions, or a new separately licensed product?