Why Cloudron's Docker only? How about VM containers with generic Docker Compose scripts?s?

[MooCloud_Matt]

Maybe I join too late on this discussion, but from what I understand the bid issue is flexibility vs cloudron ideals of Easy, Safe(backup) and it just works.

Probably expanding the Manifest specification and openly allowing the file format to be also used externally from cloudron can help its spread, and increasing the 3* party apps.

So these 2 changes could help:

• If CloudronManifest is an open standard, like Docker Compose based on the https://compose-spec.io/. Other dev could build a CLI tool to install the app based on it. And provide the community the ability to trust this format to be in the future use outside cloudron if cloudron changes its idea on supporting its community or gets sold.

• Improving the Manifest with the introduction of more than 3 directories now available, allowing the dev to set if a directory is writable and if the directory needs to be backup.
  with the list of directories, that need to be backup available on the manifest file, backups can be standardized and at the same time is easy and more convenient to containerize apps for cloudron.

[marcusquinn]

The main issues we have are:

1. We can get an app running same-day with all we need on a separate VPS, but it could be a week of trial & error and blockers with Cloudron limitations. After losing so much time already, the time and will to then try and negotiate through the blockers is diminished.
2. Many apps are multiple Docker-containers or need services that Cloudron monopolises or would just work with a VM container.
3. Why not? What really is the harm in adding this generic feature that's already there in the required Ubuntu core OS layer anyway?
4. I believe that without scaling the number of staff, packagers and developers of Cloudron, it's only a matter of time that a competitive alternative will take on these things and make Cloudron a stale relic in comparison. The app packaging bottleneck is a real bandwidth problem that seems to have slowed more over time as the developer time is split with feature priorities and maintenance. Even if the negotiation weren't extra work for convincing the gatekeepers for features or allowances, we're very, very centralised in dependancies right now, that does mean a wish list of more and more apps that have ample demand but no sign of ever being packaged at the current rate. This at-least eases that quagmire somewhat.

[fbartels]

LXC is certainly a nice tool, I am using wherever possible. But instead of starting lxc containers on a Cloudron host, I would rather do it the other way around and have one of my LXCs a Cloudron container (at some point I have to try how well Cloundron runs on ZFS).

Plus I do see challenges on the networking side, like what if you want to run an app on e.g. port 25, 110, 143, 443, 993, 995, ... Ideally you will want to have a dedicated ip per vm.

But I do agree that some sort of integration of external servers into Cloudron would be neat. Maybe a little agent sitting on external servers and establishing a private network with the Cloudron server. This agent could then expose the Cloudron usermanagement to the external server and allow easy use of Cloudron as the mail relay. Bonus points for direct integration into the Cloudron Dashboard for running webapplications. Maybe a simplified healthcheck and monitoring could be possible from the Cloudron host as well.

[MooCloud_Matt]

@marcusquinn
Partially is solvable using the additional docker container feature in Cloudron, that one to have a specific database.
I really think that most feature would be solve if we add a more open and robust manifest, like multiple volume support (in the manifest).
Maybe a script that is executed before a backup snapshot is taken, so you have the option to stop the app and dump your db.
Managing VM is difficult and to do it reliable it's better to use specific kernels modules, so i think that for that we can still use other software.
But LXC is born to be different from Docker:
LXC is 1 container == 1 OS
Docker container (still Linux container) == 1 Service/app.

[girish]

@marcusquinn said in Why Cloudron's Docker only? How about VM containers with generic Docker Compose scripts?s?:

We can get an app running same-day with all we need on a separate VPS, but it could be a week of trial & error and blockers with Cloudron limitations. After losing so much time already, the time and will to then try and negotiate through the blockers is diminished.

Just to be on the same page. So we have an external app which user somehow installed and willing to manage. The question is how we can integrate this with Cloudron?

So, to expand on https://forum.cloudron.io/topic/7076/shortcut-app what we really need is: user creates 'External app' and can choose user management as always. The icon appears but also LDAP and Email connectivity instructions that user can plug in to the external app. Maybe it can also act as a reverse proxy so that it can do healthcheck as well as manage certs. The external app only needs to have some firewall rule to allow connections from Cloudron itself.

[marcusquinn]

@girish That's all up to the user to do manually if they wish. It's just a VM Container, what happens within it is no different to what would happen on a separate VPS, we just don't want dozens of VPSs for the time and cost when we can have just one meaty server sharing resources.

Referencing https://forum.cloudron.io/topic/7076/shortcut-app might just be confusing matters.

This post is about Cloudron offering what Cockpit and Container Station offers. Kinda like Surfer - it's just a generic app for static sites and files, this VM App would be the same concept, just a blank canvas app for making VM containers.

[marcusquinn]

@MooCloud_Matt Yup, 1 OS, within which would be its own Docker instance and all other services. Run a Docker Compose script for a new app, and see what happens. Or we could just sit here hoping that some arbitrary app might get packaged before 2032 or death by too many VPSs to manage, whichever comes first.

[girish]

@marcusquinn said in Why Cloudron's Docker only? How about VM containers with generic Docker Compose scripts?s?:

It's just a VM Container, what happens within it is no different to what would happen on a separate VPS

The VM containers concept (like kata containers) works only in hypervisors/bare metal servers (which QNAP/Synology are since you purchase hardware). You cannot create VM containers in Cloud VMs.

I think this is what sysbox solves, but I don't have much experience with that.

For Cloud VMs, we are stuck with creating additional VMs for proper isolation.

[marcusquinn]

@girish My understanding is that limitation is only nesting KVM containers, which is what the cloud VPSs mostly seem to be anyway. Anecdotally, we have Proxmox running on a Hetzner Dedicated VPS and can create container VMs within that, I'm not sure why, but it's one way of cracking things, albeit without any Cloudron maintenance benefits.

[timconsidine]

@girish as I understand the thread's initial post suggestion, it is not about external apps hosted elsewhere and accessible from Cloudron dashboard.
It's about having an additional "environment" where users can more easily deploy their apps without them being in the Cloudron App Store and meeting the requirements for that.

That sounds to me like having a Portainer-like capability as a top-level app, or a Caprover-like ability to facilitate deployment of apps which are bog-standard docker run xxx or docker-compose.

As regards the latter, I've looked at self-deploying in Caprover and Dokku (self-hosted Heroku), and it's far from as simple as it seems.
Maybe it's a comfort thing but I prefer the Cloudron self-build/self-deploy approach.
So please nobody take these comments as supportive of a Caprover/Dokku facility in Cloudron.
I think it's a swamp which will suck precious resource.

A Portainer-like facility seems to me more stable and robust.

Maybe it's a question of being able to have Cloudron and Portainer running on the same (big) VPS without them affecting each other.
I'm not technical to know if that's viable.
But I think that is what is being asked for, or a way of delivering what is being asked for.

I should probably shut up now

[girish]

@timconsidine said in Why Cloudron's Docker only? How about VM containers with generic Docker Compose scripts?s?:

@girish as I understand the thread's initial post suggestion, it is not about external apps hosted elsewhere and accessible from Cloudron dashboard.

Got it I might have gotten side tracked. I guess it comes to whether we can somehow use upstream images and deploy them on Cloudron. Or atleast make packages using upstream images instead of building our own.

This is certainly worthwhile investigating. We just had a call discussing this now and we will take the top 5 upvoted apps and see if we can come up with something generic.

[girish]

@marcusquinn said in Why Cloudron's Docker only? How about VM containers with generic Docker Compose scripts?s?:

Anecdotally, we have Proxmox running on a Hetzner Dedicated VPS and can create container VMs within that, I'm not sure why

Proxmox is a hypervisor, so it can create and manage VMs.

[d19dotca]

@girish I think (my two cents anyways) part of the issue may be the lack of tutorials / concrete guides with straight-forward examples of how to package apps which is leading to an ever-growing list of app requests faster than we can package them. I know I've had a few apps I wanted to tackle packaging to learn and train myself, but the lack of resources (and lack of time on my part) is a bit of a hindrance.

So I'd like to propose one aspect of a possible solution... eventually divert some resources into coming up with a neat playground / well documented area for how to package apps and even a few example walkthroughs (i.e. how were popular apps like Invoice Ninja and WordPress packaged, Umami, etc). I know we can get through the Git and see it ourselves but I think a bit of hand-holding such as a written or video walkthrough on how it was packaged (i.e. what was the starter template, and how each step progressed) would be awesome. I know it'd help me at least and presumably others. I know some of the steps are documented and there's some template apps to work off of, but I think it could still be made easier by having some detailed walkthrough guides with popular examples.:-)

That way more people could contribute easier to the app packaging process to fit Docker-ized apps inside of Cloudron's ecosystem.

[luckow]

Oh nerds. A lot of technical thoughts

Please allow me: 2 steps back.
What is the intent of the original question? Is it a general frustration with the lack of time between an app request and a Cloudron app release (like a child waiting for Christmas)? Is the intent to have more things to play with or to compete with other apps in the same category? Is there a real need for a missing "business related" app?

Are we really missing some applications? And if so, how could we get a clear overview of the missing categories? Do we really need a third or fourth web analytics or RSS reader app in the App Store? And if so, why? IMHO, the answer should not be: because we can.

How can we find out if an app from the app wish list is worth investing time to package it as a Cloudron app with all the benefits we need as a reliable app for our daily work?

To try out apps, I have a dedicated VPS for Docker containers. I usually follow the installation instructions in the Github repository and can usually try the app after a short time. My experience is: after a short time I run into some issues where I decide that reading the announcement and playing with the app contradicts my own expectations. But sometimes I like what I get. One of my recent discoveries was Gitpod. After spending more time with Gitpod, I realized that it's not a perfect fit for Cloudron because it's very dynamic (and resource hungry) when you share the new tool with your teammates. The same goes for BigBlueButton, which is on the app wish list, but it's not worth investing time in packaging.

For me, Cloudron massively reduces my personal time spent on business critical applications. Kind of a "fire and forget." To be fair, most of the time I spend on new applications is configuring the tool, documenting it, and explaining it to my teammates. Once that's done, I forget about it until the next major release comes out, and I have to invest time to get an idea of the new features. But all that crap about updates, backups, reliability .... That's why I decided to subscribe (to pay people for their work).

Have you ever looked into a random docker.hub image? Have you ever looked into updating a random image? In my opinion, sometimes things go wrong, and sometimes they don't. So I know that mission-critical apps take time to understand, plan, and maintain. With that in mind, I've decided not to put some "cool new kid" on the app wish list. I invest time to get an idea of whether the app is worth investing time to package as a Cloudron app.

Maybe we should create a new forum category "cool new kids" where we can showcase new apps we've heard about. From there, we can invest some time (as a community) to find out if the app is worth investing time to package as a business critical app (aka Cloudron app)

[micmc]

@d19dotca said in Why Cloudron's Docker only? How about VM containers with generic Docker Compose scripts?s?:

So I'd like to propose one aspect of a possible solution... eventually divert some resources into coming up with a neat playground / well documented area for how to package apps and even a few example walkthroughs

I know it'd help me at least and presumably others. I know some of the steps are documented and there's some template apps to work off of, but I think it could still be made easier by having some detailed walkthrough guides with popular examples.:-)

That way more people could contribute easier to the app packaging process to fit Docker-ized apps inside of Cloudron's ecosystem.

I agree, from there I'd certainly contribute to package apps myself too. It's the same thing for me, the lack of time to dig deeper to figure this out with little information. I've been administrating and managing web server for 2 decades, and started to master docker and cloud technologies on top of that about 5 years back, and I'm always on the fence of getting onto try to package apps for Cloudron, however for the same reasons mentioned, when I try to get onto it then too much question pop and I've to postpone the try because of lack of time to play to figure it all.

[micmc]

@luckow said in Why Cloudron's Docker only? How about VM containers with generic Docker Compose scripts?s?:

Oh nerds. A lot of technical thoughts

Maybe we should create a new forum category "cool new kids" where we can showcase new apps we've heard about. From there, we can invest some time (as a community) to find out if the app is worth investing time to package as a business critical app (aka Cloudron app)

That's a good idea and maybe that would help put aside not only the "cool new kid" to take a look at, but also that 10th RSS reader which the new Cloudron user who just comes in, wish to have on his Cloudron because it's the one he knows and prefer. Indeed, we don't need 5 packs of each good apps out there so then we should concentrate on getting at least one good app, preferably the best one available, for each of the category we'd consider an asset to put on Cloudron that would enhance the offer.

[robi]

@girish said in Why Cloudron's Docker only? How about VM containers with generic Docker Compose scripts?s?:

I think this is what sysbox solves, but I don't have much experience with that.

Yes, it takes 10m or so to try it out.

We can circle back with @Rodny-Molina if needed.

[eddowding]

One solution might be to have a marketplace / bounty space.

1] Packages have a crowdfunded bounty behind them. Want it built? Chip in £50.

But this has a disincentive to speed, since why build for £5 when you can wait a few months for frustration to mount, and build for £500?

2] Have a month "pay per app" fee for apps which is Pay What You Want with a minimum of £1/m until a reasonable fee has been reached.

Some blend of that (see also opencollective.com) seems like it might make a viable solution.

[marcusquinn]

@girish It's a lot about speed of research before development.

As an example, we just got OpenDroneMap running the same day on a separate VPS with the Docker Compose scripts.

Doing the same thing with Cloudron would be impossible due to the multiple Docker Containers it relies upon.

With some cooperation we could package it for Cloudron - but we just don't have the week of head-banging time to add to developments, even if we can convince you to make allowances for all of this app's needs, we need to have something working and move on to the next thing.

With what I'm suggesting, we can have a VM (LXD or equivalent) container app up and running the same day, just as with any separate VPS, and we can share that.

I can't see a downside, and it's just using features that Ubuntu already offers.

We'll also help with any port-conflicts feedback or anything else that might cause issues in using the same VPS.

You have allies here, we just don't have all we need available from Cloudron, and any time we lose in battling things that won't work quickly, is time away from the additional app packaging we could be helping with for the overall cause.

[marcusquinn]

@d19dotca This only helps for apps that will work with Cloudron as-is. Our main issue is apps that require changes to Cloudron itself, they are literally impossible to package with current restrictions. I understand the restrictions, and their reasons, but they are real blockers, so we know before we have started that we will fail, so all effort then goes into separate VPSs, which is of no benefit to this ecosystem, and those VPSs can't benefit from any of this ecosystem either.