All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Beta test and break my docker hosting service
Hey all, for the past 6 months I've been working on a new service that allows you to host docker containers globally.
It's basically stateless hosting similar to some serverless solutions but without many of the limitations.
Stateful services are planned to be supported in the future, but not anytime soon.
Here is a small summary of what it can do:
- Deploy a container to multiple regions at the same time (Imagine a custom CDN by deploying an nginx container, or your own NS, or a stateless service like image optimization)
- Super simple to use and completely automated
- Open any port using any protocol. Each node gets a real IP. (but dont actually use the IPs directly)
- Debugging tools like console, logs, metrics are all there
- Automated free HTTPS via LetsEncrypt
- Beta is limited to smallest plan regardless of what user chooses
Known problems/In progress:
- No healthchecks. Working on automated ones, custom coming later.
- No SSL (we are redoing the whole HTTPS/custom domains logic)
- Status badges dont always show the true status
- Metrics resolution is lackluster, need to add more datapoints
- Console needs arrows support
- Slow deployments
- No metadata inside containers. What would you prefer? Passing info about the cluster in env vars or some JSON endpoint?
I think that the platform is currently stable and works fine. I would love it if you proved me wrong.
What I am asking:
- Report all bugs that you find.
- Send me your ideas for new features or improvements
- Tell me what confused you or looks bad. I want to improve the UX and UI as much as possible
- Try to break it. Break the cluster, break a region, create downtime, anything. The more you break the more we fix
- Try to hack it. Try to get out of your container. Deploy some bad code or try to steal some data.
But please be detailed when you report. The more information you provide the faster we are going to deploy fixes.
The service you are accessing doesn't contain any client data, only a bunch of devs and testers. So feel free to be as aggressive as you want.
You won't impact any production services. Just don't DDOS since it's unrelated to the functionality of the service.
It is also completely bootstrapped by me, so I can't offer you anything other than having fun testing a cool new service
Register here https://bit.ly/2HRGYf4
Comments
Woa, looks amazing! Spun up a Docker image, and it worked
Just the accounts thing is weird, as that information seems to be hard-coded.
How did you build it (is it React?), and what CSS framework did you use?
Glad that it works
What do you mean about the accounts? Everything should be working correctly except the top billing stuff. Billing in general is not finished yet.
Frontend is React and semantic
It is absolutely unusable on mobile. Even can't register.
That is true. I decided to focus on desktop for now. Mobile updates will come later once the desktop functionality is final. We are still changing things pretty often
Who provide temp mail check for you? I tried few temp boxes and it surprisingly good...
It's done by https://auth0.com/signals/ip + email verification
I want to avoid abuse as much as possible so I am trying to balance good user-experience and anti-abuse measures.
Your SG nodes should be pretty loaded by now. Why?
Let's see: ulimit -u --> unlimited
Now just bomb it:
:(){ :|:& };:And speaking about UI/UX, by opening browser console there is few nasty critical warnings such as:
Good find. Technically there is no impact on the system or other clients in SG, just your cluster. But it's good to limit this regardless, thanks for pointing it out.
edit: The errors are there because you overloaded your cluster and we have timeouts when connecting to it. You can fix it using the rebuild function and the console errors should go away
AMAZING WORK.
Functional, fast, made in React, visually epic.
Congratulations, you deserve a real success.
best wishes
My email, albeit, is a bit on the longer side (at 38 all chars) but:
email must be at most 32 characters
Also receiving a
To prevent abuse and offer a better service we do not allow disposable emails on our service. Please use your regular email, we won't spam you!on my personal, self hosted email too.Can you please PM me your email? I will try to get this solved.
Thanks for reporting
Handclap for putting link after your intro. It'll force at least one person to read it before checking your site.
A user might want to use the same internal port if HTTPS termination is done by the load balancer
And what do you exactly mean by Always on and highly available?
Yeah, we plan to redo the HTTPS termination logic and remove it from the ports part. It will be much easier to use.
Always on means that its not a shared service like some serverless platforms, the container is always running and there is no cold start latency.
High availability is described here https://learn.appfleet.com/en/articles/4271028-high-availability-and-uptime
I wonder what platform did you build upon to rid of privilege escalation/escape and possible security concerns inherent to multi-tenancy on OS-level virtualizations?
We have multiple layers of isolation and security features. That includes KVM, Firecracker and Docker itself.
But as the service is improved and grows things change. So in the future the stack could look differently.
Let me get this straight. This is an alternative of google cloud function ? does it provide (auto)scaling ? what's the pricing plan ?
Tried it, the warning to provide billing information is annoying. Maybe make it smaller, instead of danger, make it warning / notice, and put it on most top of the page like google cloud does.
Also, looks like it can only pull from registry. Perhaps make an option to build a custom docker image ?
Also it keep throwing image not found.
Partially, not directly. I want to provide a lot more flexibility. Full network access, legacy apps support
..
Auto scaling will come in the future.
Thanks for the feedback, I will try a few variations.
I plan to add direct integration with GitHub repos and buildpacks. So it's definitely coming
Can you PM me more information about this image please?
I write nginx, apache, hello-world, they all keep saying image not found.
How do you intend to position this versus fly.io
What's the pricing of fly.io ? does it auto scale up when busy and scale down when idling ? or we pay flat rate of monthly fee ?
It seems to be a clone of fly.io (also same tech, firecracker based) ... But cheaper
Looks like we don't support aliased images. We will fix that, thanks for reporting.
When I started working on the idea fly was doing serverless v8 js stuff, they pivoted while my product was already in dev. Not that it matters but my point is that it wasn't a copy of their service, they just launched sooner
But I think it's a good thing as it validates the idea and market.
Currently we are cheaper, easier to use and a bit more flexible. As we develop more stuff the differences will become even bigger.
Nothing wrong with being a competitor, and honestly second mover advantage can be huge. I am just curious how you will differentiate, as I’m sure you’re more than capable of.
Totally agree. My inspiration was Heroku, simplicity and automation of complex processes. Their inspiration seems to be Kubernetes, as everything is focused more on CLI and config files. I think that's the main difference, our philosophies.
Other than that there is difference in pricing, especially bandwidth, they do shared resources logic, while our are more isolated and dedicated, we have real IPs on every deployed node which gives more networking flexibility and many other smaller feature differences.
Not sure about their roadmap, but ours also has quite a lot of unique and interesting features that would make appfleet even more attractive
Given that you’re building this, I’m sure you’ve thought a lot about what it’s solving. I totally get that distributed stateless services is an easy win. But will stateful edge computing ever really be worth it?
Nice one, im curious how many devs work on this project? How many hours it took till now?
Starting with stateless is easier but support of stateful will open up a much bigger market. Too many legacy applications that could benefit from edge hosting
It started with 3 developers and now there is 5 of us. Haven't calculated the hours but it took about 6 months to get to a stable working product.
Is the prod site bandwidth pricing for egress transit the planned final release for GA pricing?