Way more NAGware, very little functionality, bugs galore

[abufrejoval]

I use Proxmox in my home-lab, mostly for fun now, since I've retired. I did want to use it to replace a bigger (Redhat) oVirt infra I was operating in a corporate lab, but that went into the cloud instead, and I left early.

So no, it's not very likely I'll buy subscriptions, now that most likely it won't ever outgrow family use and there is simply no matching license tier. And I am running it on Atoms, and aiming lower yet for some stuff: the best clusters are made from in-expensive parts, best revenues is another story, I know.

Getting nagged at logon, that's fine, as have been the levels of color and warning in the legacy GUI: it doesn' really get in the way of using it, and yes, I am considering buying every time, but that stops after a first glance at the price list.

But I can't say the same with the Datacenter Manager, where the level of nagging is killing all utiltiy: you're constantly left with a screen that is full of "critical errors", which simply doesn't reflect the true state of the nodes.

I then have been trying to find extra functionality, like better management across machines in different clusters, not clustered or a mix of both. To me clustering only really happens when nodes are independent in terms of hardware faults, which means resilient or clustered storage, ideally with managed VM migrations or automatic restarts, including respecting HA rules for application clusters (such VMs shouldn't share hosts).

In short, some of the benefits of Nutanix, VMware, oVirt etc. without their complexity, cost and lack of a future.

The only thing I was able to find was the ability to migrate VMs between hosts not in the same cluster, basically via a backup: yes, that's in fact 'automating' the three manual steps I did before. But I couldn't find more in terms of operational support, it's mostly just a monitoring screen, something we used to do in Zabbix, except that you can't escape the nags.

And when I moved some of the previously isolated machines into the cluster, I noticed that there was no way to make the manager understand that.

In fact it didn't allow me to remove a node that wasn't alive and well at all: for all I know I'd have to restart from scratch, unless there was a way to export and re-import whatever custom screens one might have made, yet without the missing nodes.

In this state of development, that's not a lot of help at all. And I am worried that this is in fact the pre-cursor to a full re-design of the GUI in Rust, with nags to stay and a focus on the huge cloud or enterprise deployments, where the hyperscalers have won and Broadcomm is squeezing VMware customers for blood.

Don't let the $$$ tempt you to leave your most loyal adopters behind, please!

And if that's the way you really want to go: let the community know, don't pull a Broadcomm++ on their --VMware when they might be committing lots of hobby- and overtime to counter the cloud squeeze.

 

[Johannes S]

If you don't like the nag screen and don't want to contribute to the development ( guess what developers need to pay bills too ) there is an easy fix: Don't use the software.
For cross-cluster migration you can also use the pct remote-migrate and qm remote-migrate command line tools. And no it's not using a backup for it.

 

[cwt]

I believe both points are valid to a certain extent.

It is entirely reasonable that Proxmox developers are compensated for their work, and subscription models are clearly appropriate for production environments. The concern many users express is not the presence of a subscription notification itself, but rather that PDM currently combines subscription warnings with actual infrastructure health states, which diminishes the clarity and usefulness of the overall system overview.

Additionally, while qm remote-migrate and pct remote-migrate are certainly valuable tools, their existence also reinforces the current criticism of PDM: at present, it appears more akin to centralized monitoring combined with migration utilities than a fully developed multi-cluster management platform.

Overall, I remain positive about the direction Proxmox is taking, and I hope that the community and smaller-scale deployments continue to remain an important consideration in its development.

 

[abufrejoval]

If you don't like the nag screen and don't want to contribute to the development ( guess what developers need to pay bills too ) there is an easy fix: Don't use the software.
For cross-cluster migration you can also use the pct remote-migrate and qm remote-migrate command line tools. And no it's not using a backup for it.

Do you honestly think that this is a valuable and thoroughly valid argument in this debate?

Or did you just shoot off against a supposed freeloader?

You should do well to remember that the Linux ecosystem isn't just people paying for support or writing code.

It's also people using the software, testing it, finding bugs, offering advice and feedback.

And the strenght of a software ecosystem (or any social code, including laws and religion) isn't alone in those who "pay" alone, but the number of adherents, collaborators or believers.

When you've grown up with open software and then take your project behind a paywall, that costs people who have invested their own and very real brain investments. You might not have made a dime out of that directly as an author or a company. But the size of the ecosystem made sure that those who paid existed and grew.

It's not outright theft, obviously, but there is a reason why it's looked upon very, very badly and can be a death spell for a software project... just as much as too little revenue.

It's not a black & white picture, but something that needs a proper balance.

That's free advice, btw. nothing I'll charge you for.

 

[meyergru]

I do not think the argument was fired against a "supposed freeloader" (I am one, too), however let's be realistic:

The concerns against PDM that you expressed boil down to two points:

1. PDM is not yet ripe enough. Granted.
2. PDM nags you more than PVE. While that is true, it is also aimed more towards business setups where one should assume that the customer is willing to pay. Fact of the matter is that most small installations do not even need and/or use cluster setups or many PVE hosts at all. If you do, you are probably the kind of customer that should pay for the software. If you don't, use Pegaprox. I would not speculate of PVE going the same route, given that we freeloaders are also free testers.

 

[abufrejoval]

If you don't like the nag screen and don't want to contribute to the development ( guess what developers need to pay bills too ) there is an easy fix: Don't use the software.
For cross-cluster migration you can also use the pct remote-migrate and qm remote-migrate command line tools. And no it's not using a backup for it

"Wehret der Anfänge" is the secondary aspect; my understanding is that the PDM isn't meant to remain a alternate tool to achieve more or less the same things the legacy UI delivered, but will ultimately replace it.

And if Proxmox is aiming at pushing out users without a subscription, they should say so today (and take the flak), not have it creep in gradually.

I also believe that the scalability shortcomings of Proxmox aren't really in the UI but at the core, so I am a bit surprised they are refactoring at that end.

 

[Johannes S]

It is entirely reasonable that Proxmox developers are compensated for their work, and subscription models are clearly appropriate for production environments. The concern many users express is not the presence of a subscription notification itself, but rather that PDM currently combines subscription warnings with actual infrastructure health states, which diminishes the clarity and usefulness of the overall system overview.

For an "overall system overview" I still think that people should use a dedicated monitoring plattform. I really don't want to have another dashboard to monitor but alerting on critical state and a general overview no matther wheter it's a vm, a PVE host or something else (like my Raspberry or OPNSense box). I don't want to have to use another monitoring software for it, no matter how great it might be. If your main reason to use the PDM is not to have to setup something like Zabbix, Prometheus or Icinga (which are all free software and don't have a nag screen) then that's fine but even such basic monitoring features needed work to be done. And imho the PDM subscription is actually cheaper than it could be: For Companys who already have a basic support subscription for their fleet the PDM is free of charge. Promox Server Solutions could also do something like: "You need at least to have a basic subscription for every node in PDM AND pay a subscription fee for PDM".

Additionally, while qm remote-migrate and pct remote-migrate are certainly valuable tools, their existence also reinforces the current criticism of PDM: at present, it appears more akin to centralized monitoring combined with migration utilities than a fully developed multi-cluster management platform.

And with this point I would agree. To change this (to have more features in PDM ) needs actual work done by developers . This needs to be funded somehow and this funding won't come from homelabbers.

 

[Johannes S]

Do you honestly think that this is a valuable and thoroughly valid argument in this debate?

Yes, otherwise I wouldn't have used it

You should do well to remember that the Linux ecosystem isn't just people paying for support or writing code.

Correct, I'm also not paying for Proxmox. But I also don't expect developers to give freebies to me. I'm trying to help in other terms, e.g. by participating in the forum or being a free beta tester And I'm not just shooting against freeloaders (although I think the entitlement of parts of the homelabbing community (especially on reddit) is quite gross) I'm also trying to help people here Whether my answers to people with issues in their setups are actually helpful is a different matter though, that's up to these people to decide.

It's also people using the software, testing it, finding bugs, offering advice and feedback.

Correct, but not every feedback is valid. And "Please give me your work for free" is an absolutely valid opinion. It's also absolutely valid for developers to ignore it.

That's free advice, btw. nothing I'll charge you for.

I'm not even part of the Proxmox team so I don't get why you address this to me. The Promox team had been around quite a long time though (more than 20 years I think?) so I trust them not to go the Broadcom way.

 

[Johannes S]

"Wehret der Anfänge" is the secondary aspect; my understanding is that the PDM isn't meant to remain a alternate tool to achieve more or less the same things the legacy UI delivered, but will ultimately replace it.

Pulling a case of Godwins law won't make me take you more seriously. In the case of Proxmox it's quite laughable since they have a decades old track record of NOT outselling. One Proxmox developer ( I think @t.lamprecht but I might be wrong) also explained their reasoning for the nag screeens (even before PDM was released) on hackernews. The link doesn't work right now so I quote the parts from an older forum discussion on the oh-so-bad-nag-screens:

In short, we're targeting enterprise users with a mix of a (soft) stick (e.g. pop-up to note that one isn't having the most production ready experience) a carrot (way better tested updates and depending on the level also enterprise support). Homelab users aren't in the target of our subscription services at all, and if we'd target them with cheap prices that would be just misused by companies too, we know this for a fact because the project is over 15 years old, and there was a lot tried out before getting to the current design. And while there might some protection mechanisms we could set up, we rather avoid DRM'ifying Proxmox projects and avoid wasting time playing cat-and-mouse games with entities trying to abuse this.

Note though, that we still cater to the homelab in other ways even if it isn't our main target audience, like being very active on our community forums for all users, or simple having 100% FLOSS software, no open-core or other, in my opinion rather questionable, open source models. And the price of a pop-up or using the still very stable no-subscription repository is IMO also quite small compared to the feature on gets 100% for free.

If one wants to contribute to our project but either cannot, or does not want, to afford a subscription, then I think helping out in the communities, submitting elaborate bug reports and thought out feature requests, spreading gospel at the companies they work with/for is not only cheaper for them, but also much more worth for the project.

> I'm not sure what "complete feature-set" means—the free offering already seems pretty complete to me, though obviously my use case is not very demanding.
I worded that part a bit oddly (sorry, not a native speaker), I meant that one already gets all features, as we don't do any feature gating or the like. I.e., you can use all features of the software with or without subscription, the latter only provides extra services.

> Still, I wonder how many homelab users would be willing to pay something like $10-20/yr (or maybe a one-time license purchase) just to make the nag dialog at login go away.

Actually the "stick" part of the carrots and stick design is just as important for enterprise users, as surprisingly a significant amount of enterprise user do not care of not getting any software updates ("it's enterprise, it just needs to magically work, even for new unforeseen events!11!!1"), and the subscriptions and enterprise repo existed before the nag, I cannot share all details, but you will have to believe me that it made a big difference. It also showed that a lot of companies are willing to pay for projects supporting their infrastructure, but they need to be made aware of the possibility quite actively.

But yeah, I can understand the opinions of the home lab community here, but I also hope that I could convey that the current system was carefully optimized over many years for our business case, and that means, that it can pay a (nowadays not so small amount of) developers salaries to extend, maintain and support the growing amounts of Proxmox projects.

If anybody that stumbles over this still does not believe in the necessity of the system, or know how it will be better if we do XYZ (how, if you have zero insights in our data and obviously do not run your own company doing this?), I just have to be blunt and recommend just using an alternative, for PVE et al. the single nag on login is the price you pay to get a full-blown cluster & hyper-visor stack.

> or maybe a one-time license purchase

I feel already like some FLOSS evangelist, but that's something I just have to correct: we sell no licenses at all, our projects are, and will stay, AGPLv3 licensed. And w.r.t. the same question for a life-time subscription with a one-time fee, not planned, reasons: see above.

btw. thanks for the discussion, most often I read this more in the demanding voice, which is hard to stay polite when responding – that wasn't the case at all here.

For the complete discussion see: https://news.ycombinator.com/item?id=39650877

In other words: The point of the nag screen is not to squeeze money out of homeusers, but to make companys who make money from using Proxmox Software contribute to their development. Imho this is only fair. While I can't afford a subscription right now for my homelab If I would be using Proxmox as a business making money I would want to have support for it and make sure, that the development of a key part of my infrastructure will continue.

I also believe that the scalability shortcomings of Proxmox aren't really in the UI but at the core, so I am a bit surprised they are refactoring at that end.

They think differently obviously and who should know better than them? I suspect a different reason though: Today it's easier to find people who can code in rust than in Perl.