pve-no-subscription or no repository at all?

[reg_ed]

I've seen a few threads here about how if one doesn't have a Proxmox subscription, the pve-no-subscription repository can be configured for updates instead.

My question is, should it be? The description of this repo is 'This is the recommended repository for testing and non-production use. Its packages are not as heavily tested and validated.' What do you consider to be 'production'? Live commercial installations only or any installation that is being used for anything other than testing?

My use case: A home user, two nodes with a handful of VMs for my 'home lab'. The minimum subscription fee is simply uneconomical for a setup as small as mine and is therefore out of the question. The description of the no-subscription repo sounds like it's full of beta packages that could have major bugs that could ruin my setup. So what's to be done?

tl;dr
What is recommended and/or safer for non-subscription setup; configure pve-no-subscription or have no PVE update repo configured at all?

 

[tom]

The description of the no-subscription repo sounds like it's full of beta packages

where did you read this? if you want beta packages, use the pvetest repo.

for home user: just use the no-subscription repo. if you run a business, go for a subscription.

 

[reg_ed]

Thank you.

 

[zodiac]

I am in a similar situation to yours, and I am actually *this* close to signing up for an annual subscription on the lowest tier. It's within budget and it's the one piece of software where I feel I can justify the cost. The thing that is holding me back is that I am still evaluating what to do about the other Proxmox products (backup, and mail). And their pricing is a lot less hobbyist-friendly. It makes the decision very confusing. Too many choices to make, and then analysis paralysis strikes. I probably should get my act together.

As for the choice of repository, it is my understanding that the no-subscription repository is a little bit akin but not identical to Debian's "unstable". You get software that is supposed to work. It was tested by Proxmox's QA team. But they can only do so much. Inevitably, once release into the wild, new and obscure bugs will pop up. Sometimes they are minor -- at other times (such as for the recent kernel update), they are surprisingly invasive.

But that's exactly why it gets released to a broad number of users who don't necessarily have the same strict uptime requirements as commercial users. Once the influx of bug reports slows down, the very same packages will then be released to subscribers, just with a small time delay. It's IMHO a fair trade-off. You either pay with money or with field testing.

 

[Dunuin]

My question is, should it be? The description of this repo is 'This is the recommended repository for testing and non-production use. Its packages are not as heavily tested and validated.' What do you consider to be 'production'?

So you prefer running a vulnerable outdated OS that never receives any important security fixes because you fear a fix could break something?

My use case: A home user, two nodes with a handful of VMs for my 'home lab'. The minimum subscription fee is simply uneconomical for a setup as small as mine and is therefore out of the question.

For me "production" is running something that is important. This could also be a homeserver where you are screwed when not being able to access important documents, you lose your password safe + the PBS with your backups of it..., virtualized router not running anymore so you might be offline for days/weeks, losing disks with important data, ...
But then again, if it is that important to you that you really rely on it, it maybe also won't hurt that much to pay 8,75€ per month per socket (so way cheaper than netflix...even with two sockets) to get the cheapest subscription for access to the enterprise repo. Only becomes problematic when running a bigger homelab with lots of small or old PVE servers so those sockets add up.

But even the no-subscription repo runs rock solid and if something isn't working usually a fix is rolling out within a few days.

 

[leesteken]

I am in a similar situation to yours, and I am actually *this* close to signing up for an annual subscription on the lowest tier. It's within budget and it's the one piece of software where I feel I can justify the cost. The thing that is holding me back is that I am still evaluating what to do about the other Proxmox products (backup, and mail). And their pricing is a lot less hobbyist-friendly.

I noticed, running Proxmox at home as a hobby for some time now, that things break less often with the enterprise repository but when they do break (and it's usually stuff like passthrough that's not guaranteed obviously), it also takes much longer to get the fix. Therefore, for my use case, I always use the no-subscription repositories.
Nonetheless, I got a no-support/community subscription anyway. I don't use it but It's just to support Proxmox. I feel that their AGPL license and support subscription model does not force you into anything, and you can get some kind of subscription to support the developers (if you can afford it, on a yearly basis).
I'm just trying to say: don't worry too much about this and don't feel you need to fit in the offered model. Just do what feels right and fits your situation.

 

[zodiac]

For me "production" is running something that is important. This could also be a homeserver where you are screwed when not being able to access important documents, you lose your password safe + the PBS with your backups of it..., virtualized router not running anymore so you might be offline for days/weeks, losing disks with important data, ...

I think I am mostly in agreement with you here. Downtime sucks. Even if it's just my homelab. It severely impacts everybody in our household and even a few hours of downtime will get me yelled at So, if things break for whatever reason, I expect having to pull an all-nighter. That's why I do think the cost of a subscription can very well be justified, even for home users. Also, this is the one piece of software that I get the most use from, but that I feel I return the least in exchange. So, paying a little bit as a "thank you" to the company is very reasonable.

On the other hand, I fully understand the point of view that all these fees add up, and if you have multi-socket systems and/or use more than one Proxmox product, then you are quickly reaching the point of diminishing returns for hobbyist usage. And honestly, even more spectacular breakage is usually something that I can fix with a few hours of work. I have administered my own Linux systems for almost three decades now. There is very little that could faze me, nor that I couldn't recover from relatively easily. So, the immediate need for the delayed-release subscription model isn't quite there.

Also, there is the appeal of getting new software as soon as it becomes available instead of waiting. It's not rational, but I do want my new "shiny"...

 

[LnxBil]

As for the choice of repository, it is my understanding that the no-subscription repository is a little bit akin but not identical to Debian's "unstable".

No, it's "testing". Debian "unstable" or SID is what PVE calls "pvetest" (overview)

I noticed, running Proxmox at home as a hobby for some time now, that things break less often with the enterprise repository but when they do break (and it's usually stuff like passthrough that's not guaranteed obviously), it also takes much longer to get the fix.

Therefore, just don't use PCIe passthrough (or Docker inside of LX(C) containers, which also break or at least broke regularly). I also tried, wasn't impressed and just got rid of it. It is the worst thing you can do with an virtualization environment because it is the opposite you try to archieve. Most of the time, you don't need it anyways. All non-testing PVE repositories are rock solid if you just do virtualization stuff (so no pcie passthrough). I've been using it for many years besides my properly-subscriptioned-production.

Also, PCIe support is part of the Ubuntu LTS kernel and every non-security-related kernel update can and probably will influence the "pcie passthrough logic" and is completely out of the scope of the Proxmox staff. It's an upstream package and problem have to be solved upstream. Therefore it may seems to take long. The problem is, that most non-enterprise users post their problems here concerning passthrough of some desktop gpu and therefore will look like it's a general problem, yet no enterprise customer will use such software. In the enterprise world, there are already products available that support virtualization out-of-the-box without any pcie passthrough shenanigans.

 

[leesteken]

Therefore, just don't use PCIe passthrough (or Docker inside of LX(C) containers, which also break or at least broke regularly). I also tried, wasn't impressed and just got rid of it. It is the worst thing you can do with an virtualization environment because it is the opposite you try to archieve. Most of the time, you don't need it anyways. All non-testing PVE repositories are rock solid if you just do virtualization stuff (so no pcie passthrough). I've been using it for many years besides my properly-subscriptioned-production.

I know its trial and error and I do think the enterprise repository is best for production. I did not intend to blame Proxmox for anything and I do see their efforts to fix things (and often quite quickly) when they happen to break accidentally. I just wanted to point out that a subscription does not force you to use any specific repository or update regime, by using my use case as an example.