Can anyone weigh in on whether any of these can be used for a cluster?
I use VMware in my homelab via vMUG, and I’m sure that’s going to get destroyed next, so I’m looking for an alternative that can allow for running VMs across hosts using shared storage with migrations between hosts. I’d prefer FOSS, but the only hypervisor I know supports all of this right now is hyper-V. I really REALLY don’t want to use hyper-v… Most of my workloads are Linux, with a handful of Windows servers that I use for an internal domain and testing.
From the brief Google searching I’ve done it appears to be possible, though, I’m not sure if proxmox skills will help me professionally. I used VMware before because I needed to learn VMware esxi and vcenter. I know it fairly well at this point.
I want to target a hypervisor solution used in large companies, I’m not sure that’s proxmox. Currently I’m leaning towards OpenStack, since I know some cloud providers use it for VPS offerings. I know enough about hyper-V that I know I don’t want to use it, ever. At least outside the context of Azure VMs. I can’t really do Azure cloud at home (they’re is a way, I’ve looked into it, but it’s very expensive), though my current workplace uses Azure extensively.
I’m just not aware of any company using proxmox as a VM platform, whether single host or clustered.
Well I can’t speak for enterprise but for me it works pretty well in a 3 node cluster. I can live transfer VMs that are hosting services with very little interruption. Proxmox also supports HA and Ceph but I haven’t used those features.
Good to know. I’ll examine everything carefully. I’ve been debating on replacing my existing monolithic iSCSI storage configuration with Ceph, so maybe that will weigh in… Having something that can access Ceph natively is a big plus. Otherwise I need something to sit in between that can basically translate Ceph to iSCSI luns, which is just more complexity that I’d like to avoid.
A lot of things to consider. Thank you for the comments.
I just don’t see us switching our 17 datacenters to proxmox. Azure HCI, perhaps, but most likely we’ll stick with vmware, at least in the foreseeable future.
The weird thing to me about the majority of VMware environments I see is that they exist to prop up and extend Microsoft environments.
Microsoft is hostile towards this use case because having your own cloud competes with their cloud products.
VMware was a commodity product that exists because they know how desperately IT professionals need to keep these Windows systems running with some level of reliability with advanced backup and replication strategies. And it was good.
After trying out proxmox I can say that:
VM performance under windows is much faster on vmware. I think this boils down to the drivers for storage. I could go more into detail but not here.
Containers and Linux VMs are offering me more than I ever really hoped for in proxmox.
But now I’m starting to think what the alternatives are really. VMware was a windows first virtualization platform. Other virtualization platforms in the open source ecosystem really put things like Linux first. Having to race to get to the point of hosting windows systems with constantly increasing licensing prices has really diminished the value to me of virtualization over all for windows.
I think we as a community need to move away from windows on the server and embrace technologies like containers,docker,podman, Kubernetes and phase out reliance on Windows.
For starters, does anybody have a rock solid setup guide for a Kubernetes Active Directory System?
Yeeahh... I'm thinking (hoping) he means an alternative LDAP/IDP, like Keycloak or Authentik..? Wanting to reduce reliance on Windows = kicking AD to the curb, too.
The problem with Samba AD in a container or Samba in container is that Samba isn’t designed to be run in a temporary environment. You could run it in a LXC container but anything beyond that will break things in the short or long term.
I figured you could get around some of the storage limitations with something like persistent volume claims. I’m testing it out at the moment. I am a big fan of LXC.
I see a few people have created docker Samba Containers and I’m giving them a whirl. Can’t say much for stability but I think it’s an interesting experiment.
I know in the past smb server didn’t work in LXC containers because certain kernel modules caused conflicts.
If you manage to create persistent containers how are you going to update them down the road? Like I have said previously, Samba isn’t designed in a way that allows for effectively hot swapping system components.
It seems like it would better to create a VM template and then setup a fail over cluster. Just make sure you have a time server somewhere on the network.
If you are dead set on containers you could try LDAP in a container. I just don’t think active directory was built for Linux containerization.
There are a few applications out there that I don’t fully understand the deployment of but seem to work in containers.
Typically the storage is mounted outside of the container and passed through in the compose file for docker. This allows your data to be persistent. Ideally you would also want those to reside in a file system that can easily be snapshot like ZFS. When you pull down a new docker container, it should just remount the same location and begin to run.
Or at least that’s how I’d imagine it would run. I feel like one would run into the same challenges people have running databases persistently in containers.
For those who don’t know, EUC stands for end user computing.
Why is so hard to setup VMs for employees? Maybe I’m missing something but it seems like a matter of just creating a virtual machine with a GPU attached.
Very significantly different performance requirements. The client communication needs tuning for fast UI response. Unified comms (zoom, teams, etc) need to be redirected to avoid bottlenecking through the server. usage patterns aren’t very well distributed (everyone logs in at 8) which means you can’t over subscribe as much.
In our case we have over 1500 employees using it, but only about 500 at a time. It’s an extreme waste of resources to have to provision 3x the hardware rather than use ephemeral systems. Also it’s much easier to patch a “gold” image and recompose entire pools than have to manage all of the systems as if they were full on laptops. Just to name a couple things off the top of my head.
Yup. That’s another reason we don’t have individual systems. And most thin clients aren’t designed to connect 1:1 to a VM. They usually need a broker of some sort.
I can relly recommend proxmox. Some years ago we switched from a 60.000€ dell VMWare Storage/Server-Setup to a three Host proxmox Setup for about half the price (to be fair, add 5-10k for Setup for our local Linux Team because we did not know much about proxmox). Mainly because we were able to place one of the Hosts in our Warehouse (connected with 10g Fiber) so there theoretically will be no harm to our production in case of water/fire/whatever in the server room because the one system can instantly take over (after some learning it works Like a Charm). I had some concerns regarding ceph, but for us it has proven Rocksolid, even while we had some real weird Switch issues it always recovered fast and without issues as soon as the connection was there. A big issue were the licensing terms for Microsoft products because with three amd-systems you have a lot of cores to buy licenses for - so we had a good excuse to substitute and cut out some products that only supported Windows environments.
I use VirtualBox right now. My daily driver windows 10 guest is so slow, that pushing the start button comes with a 20s wait. Looking at the performance monitor while this is happening, nothing pops outs as the culprit. Plenty of resources left.
I’ve always sworn to VirtualBox, but I’m going to ask my boss for a workstation pro license next time I see him.
Add comment