Reconsidering ProxMox
-
@JaredBusch said in Reconsidering ProxMox:
The only hold up I have is full backups. That is simply not practical for the real world.
I feel like it is more practical than people think. Modern "app specific" backups not withstanding, between compression and dedupe and low cost storage using full backups can work for some businesses. All, no, of course not. But when comparing Proxmox' full backups to the incrementals that we were doing before using commercial tools we are expecting (we don't have numbers on this yet, that will take months) that the cost of the additional storage will be roughly equal to the cost of the tooling that we are removing.
Not some big win, but not as bad as it sounds.
-
@StuartJordan said in Reconsidering ProxMox:
@JaredBusch I agree, I backup inside the VM. Yes not great and not as good as incremental backups from VM level.
Well, that's not necessarily true. VM level backups have no way, without in-the-VM hooks to specific applications, to back up applications. Same with in VM agents. There are nice features about blind backups from lower in the stack, but in all cases, no matter where the backup is taken from, you either need a stateless box or you need some inside agent mechanism to flush data all the way from the application to the disk.
-
@scottalanmiller said in Reconsidering ProxMox:
@Pete-S said in Reconsidering ProxMox:
@matteo-nunziati said in Reconsidering ProxMox:
Don't forget openbsd:
(https://www.openbsd.org/faq/faq16.html)That is bhyve I believe. Vmm is the kernel module of bhyve.
Ah, that would explain why I didn't know about it
Not exactly, look at this presentation from 2016 (pdf)
https://bhyvecon.org/bhyvecon2016-Mike.pdf -
Since I normally using Debian or Ubuntu LTS has a container. I discovered an issue with using fedora 31 container image. After creating the container, the network of that container will not work anymore because of the systemd version. Need to add
lxc.mount.auto: systo/etc/pve/lxc/<vm-id>.conf.Also when you use fedora 31 container image SELinux is disabled and there's no firewalld or iptables. Not exactly a big deal since you can use Proxmox firewall to manage VMs and containers.
-
@black3dynamite said in Reconsidering ProxMox:
Since I normally using Debian or Ubuntu LTS has a container. I discovered an issue with using fedora 31 container image. After creating the container, the network of that container will not work anymore because of the systemd version. Need to add
lxc.mount.auto: systo/etc/pve/lxc/<vm-id>.conf.Also when you use fedora 31 container image SELinux is disabled and there's no firewalld or iptables. Not exactly a big deal since you can use Proxmox firewall to manage VMs and containers.
Yeah that's the same with their cloud image. It's been like that for a while. They are probably assuming you're using the providers security groups to stop network access.
-
After reading Proxmox documentation, I now know why my swap is more than my RAM when using a container image.
If you assign your container 1GiB of RAM and 512 MiB of swap. The total swap is 1.5 GiB.
SWAP (1.5 GiB) = RAM (1 GiB) + SWAP (512 MiB)
-
Spun up a nested Proxmox instance.
Really liking the built-in backup & the fact that I can snapshot a UEFI vm! -
@FATeknollogee said in Reconsidering ProxMox:
Spun up a nested Proxmox instance.
Really liking the built-in backup & the fact that I can snapshot a UEFI vm!I think its only possible because of using LVM snapshot.
-
When using UEFI, it will add a EFI disk too.
-
Proxmox full backup is not bad when your lxc containers is small.
-
@black3dynamite What if you have massive containers though?
-
@DustinB3403 said in Reconsidering ProxMox:
@black3dynamite What if you have massive containers though?
I don't know since I don't use a lot of containers.
-
@DustinB3403 said in Reconsidering ProxMox:
@black3dynamite What if you have massive containers though?
Who has that? What's a use case for massive containers?
-
@stacksofplates said in Reconsidering ProxMox:
@DustinB3403 said in Reconsidering ProxMox:
@black3dynamite What if you have massive containers though?
Who has that? What's a use case for massive containers?
We do, for example. Not often, but when we need storage, containers are still better than a full VM if you don't need the extra overhead.
-
After adjusting the updates to the pve-no-subscription repo and updating (full-upgrade), I had an issue where no templates for Containers were available on the host, to "Download" (under pve -> local (pve) -> "Content" -> "Templates").
Apparently the list of available templates is updated daily through the "pve-daily-update timer".
I triggered and update manually (cause i'm impatient) via:
root@pve:~# pveam update...you can then check the list via:
root@pve:~# pveam availableAnd then I had templates to use under pve -> local (pve) -> "Content" -> "Templates".
Just spinning up a Debian 10 CT and VM now.
-
PROXMOX with LIZARD FS? ANYONE SUCCESSFULLY COMPLETED IT.
-
@mroth911 said in Reconsidering ProxMox:
PROXMOX with LIZARD FS? ANYONE SUCCESSFULLY COMPLETED IT.
I have not (but haven't tried). It doesn't appear to be a supported thing from ProxMox. They don't reference it anywhere, only LizardFS does. It's all CLI driven it looks like. Doesn't look like you get any GUI control like with Ceph.
LizardFS mentions in their documentation that they run the chunk server on the ProxMox nodes in functional testing, developemnt, and "some private setups" using LXC containers as the chunk server. So it doesn't sound like it's a recommended thing to do. I'm not sure in the other setup whether they have the chunk servers as completely separate physical machines or if it's a VM with QoS for the disks or something else. Not much out there on it.
-
We've been deploying more and more Proxmox and so far, knock on wood, we are remaining happy.
-
I'm quite happy with it as well, I've deployed a couple of production servers using proxmox and all good so far as well
-
@scottalanmiller said in Reconsidering ProxMox:
We've been deploying more and more Proxmox and so far, knock on wood, we are remaining happy.
Are you deploying multi-node setups or just single nodes?
