28 ms·

Why companies move off Heroku (besides the cost)

275 points by jusrhee 4y ago

jmuguy 4y ago
Not having any sort of basic disk/storage solution has been painful for us. There's been a lot of situations where I would love to store a big blob of data on the disk while processing it but if that Heroku dyno reboots - poof, its gone.
- jrvarela56 4y ago
  What kind of workflow are you trying to fit? My impression is that this lack of 'instance storage' is by design, as well as other constraints that can be mapped to these: https://12factor.net/ https://12factor.net/
  - jmuguy 4y ago
    I imagine Heroku would say the same, and thats fine. We deploy a bunch of Rails apps, backed by Postgres, via Heroku and most of the time what we're doing fits neatly into the 12 factor definition. However what happens if you're doing something that doesn't really fit that definition. For instance we have a Rails app with some background workers doing data processing. I would very much like to have these workers just dump to disk, so I can take a big chunk of data and move it into Postgres at once. But I can't do that with Heroku. So basically this is something that Heroku isn't designed to do, but its also something I would rather not need to go one level of abstraction deeper, to AWS directly for instance, in order to do. And its also something that every single one of their competitors offers.
    - jrvarela56 4y ago
      My hunch is that if you try to build your solution with the constraints enforced by Heroku your system will be more resilient. If it's about being easy to dump stuff to disk, using S3 or even Postgres for something like this could become easy after a few attempts. Another pattern that would come out of this exercise is trying to deal with that job in parallel by breaking it up into sub-tasks (many workers at the same time). Anyway, too little context for me to flesh out a solution but the gist is that in general these constraints nudge you in a 'healthy' direction when designing systems.
- mrkurt 4y ago
  Boy howdy do I agree with this. Filesystems are immensely handy. We (a sometimes competitor of Heroku) shipped persistent volumes very early and it's been amazingly empowering. The DX gets difficult, but we're ironing that out over time
hpkuarg 4y ago
On the subject of cost, I am reminded of Pinboard's tweet[0]. [0]: https://twitter.com/pinboard/status/494239199038828544 https://twitter.com/pinboard/status/494239199038828544
baggiponte 4y ago
Alternatives to Heroku besides Porter?
- focom 4y ago
  dokku on a vps? ref: https://github.com/dokku/dokku https://github.com/dokku/dokku
- sascha_sl 4y ago
  fly.io is promising in the direct competitor space Dokku is a good alternative for running herokuish on a single server with lots of control.
  - josegonzalez 4y ago
    Dokku Maintainer here. Dokku also supports Dockerfiles, Docker Images, Tarballs (similar to heroku slugs), and Cloud Native Buildpacks. I'm also actively working on AWS Lambda support[1] (both for simple usage without much config as well as SAM-based usage) and investigating Replicate's Cog[2] and Railways Nixpacks[3] functionalities for building apps. There are quite a few options in the OSS space (as well as Commercial offerings from new startups and popular incumbents). It's an interesting space to be in, and its always fun to see how new offerings innovate on existing solutions. [1] https://github.com/dokku/lambda-builder [2] https://github.com/replicate/cog [3] https://github.com/railwayapp/nixpacks
- ywain 4y ago
  Render (https://render.com/ https://render.com/) is pretty nice.
- hugg 4y ago
  we're migrating to something called cloudflare pages, but i'm not involved in that
- thibaut_barrere 4y ago
  A couple of French-based options that I've been using as replacements: - https://www.clever-cloud.com https://www.clever-cloud.com - https://scalingo.com https://scalingo.com - https://www.qovery.com https://www.qovery.com
soperj 4y ago
Begs the question, what do people not like about Porter?
chimen 4y ago
Why companies move on to Heroku should be the question. All raised issues are there since Heroku.
rememberlenny 4y ago
Porter customer support is insanely good. If you are trying to migrate from Heroku to AWS, they make it incredibly easy and go out of their way to engineer alongside you. I was deeply impressed as a user and have nothing but positive things to say about their team.
craigkerstiens 4y ago
As someone that was there at Heroku after the acquisition, I don't think you can state that since the acquisition it's been consistently downhill. There isn't much debate that things have stagnated in the last 5-7 years which is a longer story on why probably. Some examples of innovation that happened and were launched after the acquisition: buildpacks (at the time of acquisition Heroku was still Ruby only), Heroku Postgres launched forks/followers/dataclips all after the acquisition, review apps came several years after. Salesforce may have had an eventual hand in it, but there was still a lot of innovation happening due to the folks there in the near to mid-term after the acquisition for several years. All that said, very excited for the new crop of players in the space. There are a number of companies trying to be a cheaper or more stable Heroku. Personally I'm excited about the ones that are taking their own unique approach. https://www.fly.io https://www.fly.io and https://www.railway.app https://www.railway.app are two that to me seem to bring their own perspective vs. just trying to recreate Heroku as a carbon copy clone. There are a number more in the jamstack space that have become staples such as Netlify and Vercel which are also doing great things.
- booi 4y ago
  https://render.com https://render.com is my favorite in this space
  - bravura 4y ago
    Care to explain why you like these choices?
    - kjksf 4y ago
      I can only speak about render.com as I use for a mix of 10 static websites / Go webapps + cloudflare for DNS and caching proxy. It sounds prosaic but "it just works". Specifically compared to Digital Ocean Apps (which I used before render): * the dashboard UI is better designed and faster * the builds and deployments happen faster * similar price (to Digital Ocean, much cheaper than Heroku) * similar capabilities (attached disks, hosted Postgres, hosted redis) but the velocity seems better i.e. render.com seems to implement features faster than DO You would think that "well designed dashboard that displays instantly" would be a table stakes in an offering like that, but sadly it isn't.
jimmar 4y ago
I took a few minutes to scan Porter's website. They have a free tier. But when I read the getting started documentation, it says that I have to provision everything in my existing cloud infrastructure (e.g.; AWS). So I'm paying for the services in AWS. It seems I got lost in the story. Porter doesn't immediately feel like a replacement for Heroku.
- jusrhee 4y ago
  OP and Porter founder here. The article was meant to outline the most common technical limitations we see companies on Heroku bump up against as they outgrow Heroku. For individuals and teams running smaller workloads on Heroku where saving $ is a chief concern, Heroku is probably still a good option even though they’re declining in market share (this unprecedented recent outage aside). Porter is designed for companies that are maturing off Heroku for the technical reasons we mention or for those already looking to get the automation of Heroku in their own AWS/GCP cloud.
- asenchi 4y ago
  Make sure you are careful of their LICENSE if you choose to self-host. That repo is scary from a business perspective.
gkoberger 4y ago
I'm historically a huge Heroku defender. Been using it for a decade now, and my company still uses it despite being quite large and getting a lot of traffic. It's always been a great product and the early Salesforce days they DID ship a ton of new stuff and improve rapidly (despite a narrative that they didn't). Like, it got REALLY GOOD the years after acquisition. That being said... it's insane that we haven't been able to deploy for over two weeks and nobody there seems to care. So we're looking to move now, since it's clear Heroku has pretty much just given up at this point.
- krallja 4y ago
  > we haven't been able to deploy for over two weeks What is stopping you? Are you unable to type `git push heroku master`?
  - tommoor 4y ago
    That's fine as a solo developer – but what about a team with CI pipelines etc?
    - VWWHFSfQ 4y ago
      can't you put that in the ci pipeline
      - scrose 4y ago
        I’m pretty confident if ‘putting it in the ci pipeline’ were a straight-forward option for most people, they probably wouldn’t be paying Heroku to manage review apps. I’ve used Heroku review apps for years and have also written and taken over different custom deployment pipelines. Review apps have a million different ways to be a giant time and money sink if not planned and implemented properly.
  - gkoberger 4y ago
    I think you're underestimating the complexity of our codebase. Aside from the usual blockers, we have hundreds of repos (our Staging feature for Enterprise allows customers to deploy our code on their own cadence) all tied together using pipelines. We have our own custom release management software, which now doesn't work. Different repos have to go out at the same time so things don't break. Plus, we extensively use their review apps for code reviews, which we've lost access to. Lastly, not everyone has access to deploy directly to Heroku, so not everyone would be able to 'git push heroku main'. Could we fix all of this and get it working? Yeah. But we want to be focusing on building our product, which is why we pay Heroku a ton of money so we don't have to worry about this.
dml2135 4y ago
I'm at a small startup (about 8 engineers) and for the most part Heroku has been a great solution for a team of our size. However, the recent github incident was a major PITA for us, and has us seriously considering leaving the platform for the first time. We're still working on getting all of our automated deployments back up and running. In addition, we had just put in some work to start using the Review Apps feature, which now seems to be gone for the foreseeable future.
- fomb 4y ago
  Github actions are your friend friend: https://github.com/AkhileshNS/heroku-deploy https://github.com/AkhileshNS/heroku-deploy https://github.com/pmbanugo/heroku-review-app-actions https://github.com/pmbanugo/heroku-review-app-actions
  - dml2135 4y ago
    This is awesome, thanks!
vmception 4y ago
Netlify do everything that I had used Heroku for, for free. In nearly one click too, instead of a dozen levers.
- ajhurliman 4y ago
  Do you only use lambdas for compute, or do they host servers now?
  - vmception 4y ago
    They can have backends too, not just stuff rendered for the client You can also have those discreet compute instances as well, Netlify I think just recently introduced this, not sure how recent To my knowledge the only thing missing is a persistent storage solution, so you still need a database elsewhere but your api can read and write to it I personally dont do system design that way with my “web3” sites, as these are basically frontends to smart contracts, so I’m essentially using nearby nodes as compute instances and users pay to update the state of the compute instances and what the client therefore displays. Surprisingly economical for developers, I see that model attracting more devs very quickly as the stack and deployment is less complicated and cheaper and customers are already there.
intothemild 4y ago
Other than Dokku how many other PaaS solutions can be run on your own hardware?
- c17r 4y ago
  - piku https://github.com/piku/piku https://github.com/piku/piku - caprover https://caprover.com/ https://caprover.com/
- jnettome 4y ago
  we have been using caprover for 6 months, until we got the first overload and I had to find what exactly was happening during an outage. I love the solution but decided to "go by myself" after this. currently using a combination of docker swarm, traefik (via service labels) and portainer. totally in love, doing good for our needs (1 monolith, 12 microservices)
arberx 4y ago
One thing this article is missing is the arbitrary "400MB slug size" that heroku limits your application to if you want to deploy.
- jacobsenscott 4y ago
  FWIW that's a "soft limit". I've deployed slugs over that limit without trouble. Idk if they have a hard limit.
  - henryaj 4y ago
    Hard limit is 500MB.
    - fomb 4y ago
      Which isn't hard. Support can increase it, but there are some proviso's such as increased deploy/boot time
davidkuennen 4y ago
I don't know if I'm seeing things, but Heroku seems to be more expensive than GCP. I get a 8 GB RAM Postgres instance for 200$ on Heroku. On GCP I get a 4CPU 26GB RAM Postgres instance for 280$ (Cloud SQL). What's the USP Heroku provides?
- RedShift1 4y ago
  They were early in the devops game, and provided an easy way to run things without having to deal with hardware, networks, scaling, operating systems, maintenance, etc... yourself. Nowadays I think they are no longer running on innovation, but on inertia.
- nicoburns 4y ago
  Point it at your got repo, click to choose which tech stack you’re running, write a 1 line yaml file to tell it what command to run to start up, and you already have a server running and a build pipeline that deploys when you push your code.
crcastle 4y ago
Small nit: I believe this is incorrect: > More recently, all Git-based deployments (which is to say, virtually all deployments) to Heroku were blocked and review apps were halted for all users as a result of a GitHub OAuth token leak. It should read "all GitHub-based deployments". You can still deploy with `git push heroku main`.
- jusrhee 4y ago
  Ah yup good catch - just updated the article
  - nomilk 4y ago
    It caught my eye too, but for a different reason, this bit doesn't seem right: > which is to say, virtually all deployments My understanding is if deploying with `git push heroku main`, that application's GitHub repository was not viewable by hackers (but those apps deployed through 'Heroku GitHub Deploys' were). (please tell me if my understanding is incorrect). I think most Heroku users would deploy with `git push heroku main`, although that's purely hunch. Unrelated, but I'd add one more thing to the article, which is that Heroku docs aren't easy to give feedback on. I'd love for the docs to be on GitHub so shortcomings or inaccuracies can quickly be addressed. Currently, to point out a correction to the docs, you'd have to write a support ticket and 100% chance that support ticket isn't going beyond the person who received it, so nothing will get actioned.
justjake 4y ago
Warning: There are a couple people in this thread mentioning products that they're investors in. I won't call them out expressly, but I think people should be direct about their incentive alignments. It's disingenuous and borders on astroturfing IMO. Background: Founder of Railway.app here. There's a lot of these companies popping up that offer a "Heroku replacement", and once you dive in, you realize you have to pay $300/mo for a Kubecluster + $200/mo for the wrapped Kube service In our experience, people move off Heroku for a couple things: - Cost: Kinda self explanatory but Heroku pricing ramps hard - Flexibility: Heroku's not great for anything beyond stateful monoliths - Scalability: Notoriously Heroku's SLAs aren't that great In my mind, you don't replace Heroku with a minimum $500/mo Kubernetes cluster. Not only is this cost prohibitively expensive, but Kubernetes itself is a jet engine, and if you're not trained to use it correctly, you can risk catastrophic failure (on costs ballooning, on dataloss, etc) We're working hard at Railway to provide not just a Heroku replacement, but a next generation, composable infrastructure canvas. $5/mo, 30 seconds, and you're up and running. Demo:https://user-images.githubusercontent.com/5499880/165187948-8127b299-de74-49bb-81d4-c2e57ba66833.mp4 https://user-images.githubusercontent.com/5499880/165187948-... Would love any feedback and thoughts people have about Heroku, my thoughts above, and what we're building :) EDIT: Wow, just went from 11 -> 8 upvotes. I suppose they didn't like the astroturfing callout EDIT2: Render isn't who I'm talking about. They're great.
- heartbreak 4y ago
  They’ve been shilling their R•nder and P•rter services on every single thread about this Heroku incident, and it’s frankly a problem that HN moderators should address.
  - ryanSrich 4y ago
    What should HN do? People are free to downvote it until it's dead. That seems a more democratic way to handle it.
    - justjake 4y ago
      I'm such a HN normie does HN even have a downvote button? I can't find it
rubyist5eva 4y ago
So this is more of a sales pitch for Porter - which is fine, I didn't know it existed and these are pretty compelling reasons to try it instead of Heroku. For me - the real selling point is that yes, Heroku is expensive - but the cost of a competent devops engineer is much, much, much higher. So unless it's literally impossible to do something with Heroku that needs a specialized skillset - the "devops as a service" part of Heroku is worth every penny in my opinion.
- ajhurliman 4y ago
  Agreed. We have a small number of high value users so we’re actually running on the smallest non-free tier (mostly to avoid cold starts) with almost zero head aches. It’s like $15 a month. The only frustration we’ve hit was mentioned in the article: static IP for integration with 3rd party services (VGS), but we used an add-on and I think we’re still in the free tier, or it’s a few dollars a month. The cost? We were actually spending more on GCP due to ham fisted provisioning, plus a $150k dev ops guy. So ya, I’m pretty happy about the cost.
paxys 4y ago
This is just an ad for their own service. Heroku is doing fine. They provide exactly what they advertise, and the service has been rock solid for over a decade. Sure it doesn't fit everyone's use cases, but it doesn't need to. Ultimately platform-as-a-service has always been a dead end. Companies either want more control over the infrastructure (so use plain VMs or Kubernetes), or want to forget about servers entirely (opting for services like Lambda and now edge computing). Everything in the middle (Elastic Beanstalk, App Engine, Heroku) has been stagnant for a long time now.
- jacobsenscott 4y ago
  Taking over two weeks to fix their GH integration is not "doing fine.". I've been a huge Heroku booster for years, but you can't ignore this. The GH breakage is a huge red flag.
jacobsenscott 4y ago
For all heroku's frustrations (and I agree with all of them in the article), it is still the only thing that "just works" for a standard monolith web app. Heroku is not cheap, but it is still cheaper than a couple full time employees + aws. It is really too bad they aren't innovating - they are just burning up their 10 year lead in the space. I wouldn't start a new project on Heroku - not because of the cost, but because I don't expect them to last another 10 years.
- spicybright 4y ago
  It's honestly strange how no one has eaten their lunch yet. Besides the extra services they now offer (which you aren't required for hosting a simple app), you could probably have a team bang out the basics in a few months. You could even just be a layer on top of AWS and probably make profit from not many users, as long as you're cheaper.
  - mixedCase 4y ago
    > You could even just be a layer on top of AWS and probably make profit from not many users, as long as you're cheaper. Isn't Heroku exactly that?
  - jacobsenscott 4y ago
    There are a lot of startups that have banged out the basics in a few months. The are getting mentioned all over in the threads under this article. But the basics aren't good enough. Heroku looks simple, but does a lot, and pretty much does it all right. Their documentation alone is amazing and it would take more than a few months to create docs of the same quality.
    - rsanheim 4y ago
      Yuuup. That is the genius of Heroku. They make almost everything very simple and the tougher stuff relatively easy to understand and configure. All of that took many many years of iteration and refinement, which is why the competitors are going to take awhile to get there.
mark_l_watson 4y ago
A good article, except for one big thing: this is not an end user, rather a direct competitor. Public articles that criticize competitors always run me the wrong way. I haven't used Heroku in a few years, but it has served (using the Hobby plan) as a really low cost way to host web apps. I have been reading through the comments on alternative providers, and even though I haven't used it, GCP's Cloud Deploy looks interesting also (a very long time ago, I used AppEngine a fair amount).
- vmception 4y ago
  Check Netlify Or Vercel Best scaling in their free plans Heroku hobby is a joke in comparison and hasnt been updated in a decade while all their addons have gotten less and less featured while costing more and more I host all my static assets on IPFS which practically nullifies the bandwidth limits of Netlify
  - lambdaba 4y ago
    Could you elaborate on how you use IPFS? How fast is it? Do you use use a pinning service?
    - vmception 4y ago
      I upload static assets to IPFS and add those static links to my project instead of having an "assets" folder. IPFS is fast enough, I had these concerns too so I use a gateway for links as some of them cache files on their own CDN, for free. Yes, I use a pinning service, I was extremely hesitant as many of them are just SaaS hosting like Pinata is and charge by bandwidth used, but there are free ones like https://web3.storage https://web3.storage which for some reason is combining Filecoin and IPFS together for each upload and liveness/pinning. Ironically, Pinata also acts as a gateway and can then be used for free if you want. But there are other gateways. Cloudfront has one but what it decides to cache is strange.
- krthr 4y ago
  GCP Cloud Build + Cloud Run is a pretty good alternative for Heroku
digianarchist 4y ago
The year is 2022 and still no HTTP2 support. https://help.heroku.com/JAOCNZ25/does-heroku-have-plans-to-support-http-2 https://help.heroku.com/JAOCNZ25/does-heroku-have-plans-to-s...
- multiplegeorges 4y ago
  Or automated SSL for wildcard domains. > Supporting wildcard domains is something we'd like to do. Up until March 13, 2018, this was an upstream limitation from Let's Encrypt. ... 2018.
- Beltalowda 4y ago
  The changelog that's linked in there is depressing: only updates for languages/environments and an occasional minor update to the actual platform every few months.
- jacobsenscott 4y ago
  Yes, this is :(, but you should be serving all your static assets through a CDN, which will have http2. I can't think of a scenario where I would want to serve anything other than the dynamic content off the heroku server. I don't think http2 buys you much there. So you should be able to get away with at most 2 requests - one to the heroku server, and one to the cdn.
  - mrkurt 4y ago
    The 2022 answer is actually to bake the CDN into your PaaS. Requiring a whole separate CDN for static assets is complication no one needs in their life.
  - bpicolo 4y ago
    Need http/2 for grpc, which is certainly gaining popularity these days
intrepidsoldier 4y ago
In the last decade, a vast majority of DevTools companies have positioned themselves as “A better Heroku.” And yet, when people actually need a Heroku alternative, everything comes up short. This is a testament to a solid product Heroku built. It was definitely ahead of its time.
- leetrout 4y ago
  Bingo. Porter is one of the only ones that seems to offer feature parity across aws and gcp. Render.com looks cool but is also AWS only.
  - anurag 4y ago
    (Render employee) We currently use AWS, GCP, and Cloudflare under the hood.
    - leetrout 4y ago
      Thanks for clarifying!
zoomzoom 4y ago
First off, I would strongly disagree with the idea that Heroku post-acquisition was stagnant. As well-said elsewhere, tons of product was shipped to industry-best standards: postgres/dataclips, Heroku Connect, buildpacks - as well as introduction of CI and preview apps. But it's clear that in the post-k8s and post infra-as-code world there's been a surge in new options for how teams manage their infra and DevOps toolchain. Of course, I’d agree with the idea that conflicts of interest should be disclaimed: I’m the co-founder of Coherence (https://www.withcoherence.com https://www.withcoherence.com). Seems the key issues raised by the article and comments here are: - costs and resource control constraints when using Heroku - fixed base costs of k8s when using self-hosted PaaS type systems - flexibility constraints of heroku-similar PaaS - the desire for teams to get more ownership than Heroku gives them when it comes to configurability, reliability & security Coherence disagrees that the next generation of PaaS should be a black-box - like Heroku, many next-gen PaaS are not hosted in your own cloud and re-implement the wheel when it comes to functionality like persistent storage, hosted databases, and application support services like redis. In the end, we believe that building on top of the major clouds (Google, AWS, Azure, CloudFlare, etc…) is the right choice. It’s also important to us that you're able to host customer data in mature systems that you control. This philosophy also allows Coherence to help you use managed services the large providers have built, like Cloud Run or App Runner, which at least partially mitigate the cost and complexity risks of k8s. Coherence is vertically integrated, composable, and opinionated, with a focus on developer experience. A defined workflow for production-quality full-stack web apps with dev and production built in alongside automated test environments, including CI/CD and cloud IDEs - all configured with one high-level YAML. We’re in a very early private beta on google cloud right now - if you’re interested, please check out our site above and let us know!
- jeremyjh 4y ago
  I was just looking at AWS App Runner today and just thinking about how miserable it would be to figure out how to use AWS's proprietary build pipeline tools to replicate what I have with Heroku - and I've been using AWS forever - I just hate their UX, their docs, really every thing they do except the actual infrastructure services. So I was just thinking someone should build a product that gives me the Heroku CLI, pipelines, add-ons, review apps etc but implements it all on top of standard AWS services. No runtime services at all, just orchestrating configuration and metadata, service provisioning etc. I know there are other companies in this space but I wish someone would just blatantly steal the Heroku UX and then sell me that.
tills13 4y ago
This seems like more of an ad / hit-piece than a discussion piece.
aantix 4y ago
I've heard good things about using Hatchbox to deploy to AWS. https://www.hatchbox.io/ https://www.hatchbox.io/
bradleybuda 4y ago
We're big Heroku customers; despite that, I largely agree with the points in this article and there's a voice in the back of my head asking every few months "is this worth it?" Whenever I research the new crop of Heroku clones (the one being hawked here, and others) the pitch is always "it's just like Heroku but you can run it in your own cloud". It's mind-boggling to me that none of the clones understands that I DON'T WANT TO RUN IT. Yes, I pay Heroku a premium because I like their software (the pipelines are great, dyno formations mostly Just Work) but what I'm really paying for is: * Never typing "ssh" * Never thinking about a full disk from a runaway log file * Never thinking about a load balancer or a certificate * Never waking up because a Postgres host has failed * etc, etc I have no interest in a "Heroku but you run it" PaaS but I'd pay though the nose for a "Heroku but it's actively developed" PaaS.
- dontlaugh 4y ago
  AWS’s ECS Fargate with Aurora Serverless and AWS CA certificates is close.
  - ketzo 4y ago
    But even that stuff is gonna be more configuration than Heroku, right? I wanna spend literally zero hours a month on server / dev ops, if I can manage it. I will pay for it and accept the constraints of simplicity.
    - vlunkr 4y ago
      You can invest more time up front and automate it with an IAC tool and and CI tool to build/deploy. Probably nothing will be as hands off as Heroku, but there are lots of measures that can get you closer.
    - maccard 4y ago
      It is more configuration, yes, but it's far less than say ec2 and nomad. We run a very light production load on it and in 6 months I've had to intervene once (to bump our limits because we spiked slightly more than I expected us to)
wizwit999 4y ago
CDK +AWS is the way to go.
senjin 4y ago
Throwing this out there since it hasn't been mentioned yet but AWS's copilot tool (https://aws.github.io/copilot-cli/ https://aws.github.io/copilot-cli/) has been really nice for setting up an ECS cluster with a good build pipeline that we use in almost the same way as we use Heroku. It's definitely more mental overhead than Heroku but it offers more flexibility and tying directly to your AWS account. After setup it has been maintenance free for over a year but debugging parts of the build pipeline were painful at first. It recently started supporting AppRunner now that AppRunner can access VPCs so I'm looking forward to trying that at some point too.
chasd00 4y ago
Linode + dokku works pretty well for me
gwbas1c 4y ago
Awhile ago, my hobby Heroku site was shut down for vauge "violation of policy" reasons. Turns out it was an automated bot. When they brought it back up they kept breaking this and that... I kept having to re-open tickets. I'd really like a free alternative. (Or maybe I should figure out how to host from my NAS.)
caseydm 4y ago
Heroku is my tried and true. I've likely done thousands of deploys over the years and have never had one go bad or leave my app in a bad state. I have rolled back an app a few times due to bad configs and it was always quick and left me with a huge sigh of relief. That being said, they are very stagnant on features. A dyno has operated the same way, at the same size, at the same price for like 10 years now? The main features I would like to see are auto-scaling for all dynos and IPs that allow you to restrict database traffic. Render looks promising in that department and could likely get business from me if Heroku doesn't change in the next couple years.
rsanek 4y ago
If anything, this article solidified for me that the main reason remains cost only. Points (3) and (4) are probably not among the most important things for the vast majority of co's, even if they have 10s or 100s of employees.
ksajadi 4y ago
https://cloud66.com https://cloud66.com is the best alternative that runs on your own cloud and is native for Rails apps
antihero 4y ago
Have people that have grown out of heroku considered more dull but battle-tested solutions like CloudFoundry where you run a buildpack driven heroku-like abstraction on top of your own cloud provider (or whatever you want really)? Very widely used by enterprise, but also free and open-source.
ab-dm 4y ago
I built my first business on Heroku. Started 5 years ago and still, for the most part, it's going strong. There are some really strong downsides (especially after the last 2 weeks) but as a dev team of 1 (mostly) it's been a life saver, and that's enough to overrule all the other downsides. Having said that, we've suffered our own issues, with cost, latency (we're in AU) and scaling being the main ones. We're building a new product that we're betting very heavily on and it's all on AWS serverless. I wouldn't choose Heroku again. I think it was worth it 5 years ago, but as time goes on it becomes harder to justify.
pantulis 4y ago
key thing for me here is: why did Salesforce buy Heroku? How does Heroku fit into Salesforce strategy? When Heroku was incorporated by Salesforce there was no consumer facing capabilities in SF: everything was for business/sales users so Heroku was attractive to build public facing assets seamlessly connected to SF data (hence Heroku Connect). But then SF started incoporating other consumer facing featuers: Communities, Commerce (Demandware), heck they even have their own CMS now. One can predict they will push lowcode/citizen development beyond Force.com. Where does that leave Heroku? What does it bring to the table now for Salesforce?
rurban 4y ago
Losing GH automation did cost me exactly 20min. changing the alias from push github to push to both was trivial. I think Heroku did the right thing, but Travis still hasn't responded and changed anything, whilst the exploitation is still going on. If only I got get rid of Travis in my GitHub integrations. Deleting the app and actions and hooks hadn't changed anything. Travis has no access anymore, but PR's still block on the not existing Travis CI.
93po 4y ago
Heroku disabled my entire account with zero notice due to a fraudulant DMCA takedown notice, costing me tens of thousands of dollars. I struggled to be able to speak to anyone, and when I finally got a hold of someone, they treated me like a criminal despite making it clear I was the victim. Really shitty attitude, and the end result was getting my account reinstated but only with a "don't let this happen again, shithead" sort of treatment.
kasia66 4y ago
Usually, companies migrate from Heroku to cut the cost as they grow, but there are other reasons too: - The recent Heroku outages. - Lack of flexibility to adjust the available CPU, as Heroku offers six basic dyno types. As a result when the company grows they need to deal with overhears. - Heroku runs its servers on AWS, however, developers do not have access or control over the regions. This is an issue for the companies that deal with data requirements. - It does not offer support for crons/jobs. - Heroku default environment does not offer static IPs. (Private Spaces are only available on the Heroku Enterprise subscription.) Have a look at cloud66.com, it creates an environment like Heroku but on your servers on any cloud. This creates many benefits, including persistent storage and support for all available regions of your cloud provider of choice. But it also makes a big difference in availability: your application is not dependent on Cloud 66’s availability and won’t go down. Disclaimer:I work at Cloud 66.
cakoose 4y ago
Started something new last year and was considering Heroku, but their HA database offering uses async replication: https://devcenter.heroku.com/articles/heroku-postgres-ha#failover-conditions https://devcenter.heroku.com/articles/heroku-postgres-ha#fai... > Like followers, standbys are kept up to date asynchronously. This means that it is possible for data to be committed on the primary database but not yet on the standby. In order to minimize data loss we take two very important steps: > 1. We do not attempt the failover if the standby is more than 10 segments behind. This means the maximum possible loss is 160MB or 10 minutes, whichever is less. > 2. If any of the 10 segments were successfully archived through continuous protection, but not applied during the two minute confirmation period, we make sure they are applied before bringing the standby out of read-only mode. > Typically there is little loss to committed data. AWS RDS and GCP Cloud SQL do synchronous replication, so you are much less likely to lose data in common hardware failure scenarios. With those options available, a managed DB with async replication is a no-go. I don't understand how so many businesses would be ok with it. (I suspect most people don't even realize that Heroku HA is async.)

275 points by jusrhee 4y ago

Not having any sort of basic disk/storage solution has been painful for us. There's been a lot of situations where I would love to store a big blob of data on the disk while processing it but if that Heroku dyno reboots - poof, its gone.

jrvarela56 4y ago
What kind of workflow are you trying to fit? My impression is that this lack of 'instance storage' is by design, as well as other constraints that can be mapped to these: https://12factor.net/ https://12factor.net/
- jmuguy 4y ago
  I imagine Heroku would say the same, and thats fine. We deploy a bunch of Rails apps, backed by Postgres, via Heroku and most of the time what we're doing fits neatly into the 12 factor definition. However what happens if you're doing something that doesn't really fit that definition. For instance we have a Rails app with some background workers doing data processing. I would very much like to have these workers just dump to disk, so I can take a big chunk of data and move it into Postgres at once. But I can't do that with Heroku. So basically this is something that Heroku isn't designed to do, but its also something I would rather not need to go one level of abstraction deeper, to AWS directly for instance, in order to do. And its also something that every single one of their competitors offers.
  - jrvarela56 4y ago
    My hunch is that if you try to build your solution with the constraints enforced by Heroku your system will be more resilient. If it's about being easy to dump stuff to disk, using S3 or even Postgres for something like this could become easy after a few attempts. Another pattern that would come out of this exercise is trying to deal with that job in parallel by breaking it up into sub-tasks (many workers at the same time). Anyway, too little context for me to flesh out a solution but the gist is that in general these constraints nudge you in a 'healthy' direction when designing systems.
mrkurt 4y ago
Boy howdy do I agree with this. Filesystems are immensely handy. We (a sometimes competitor of Heroku) shipped persistent volumes very early and it's been amazingly empowering. The DX gets difficult, but we're ironing that out over time

hpkuarg 4y ago

On the subject of cost, I am reminded of Pinboard's tweet[0]. [0]: https://twitter.com/pinboard/status/494239199038828544 https://twitter.com/pinboard/status/494239199038828544

baggiponte 4y ago

Alternatives to Heroku besides Porter?

focom 4y ago
dokku on a vps? ref: https://github.com/dokku/dokku https://github.com/dokku/dokku
sascha_sl 4y ago
fly.io is promising in the direct competitor space Dokku is a good alternative for running herokuish on a single server with lots of control.
- josegonzalez 4y ago
  Dokku Maintainer here. Dokku also supports Dockerfiles, Docker Images, Tarballs (similar to heroku slugs), and Cloud Native Buildpacks. I'm also actively working on AWS Lambda support[1] (both for simple usage without much config as well as SAM-based usage) and investigating Replicate's Cog[2] and Railways Nixpacks[3] functionalities for building apps. There are quite a few options in the OSS space (as well as Commercial offerings from new startups and popular incumbents). It's an interesting space to be in, and its always fun to see how new offerings innovate on existing solutions. [1] https://github.com/dokku/lambda-builder [2] https://github.com/replicate/cog [3] https://github.com/railwayapp/nixpacks
ywain 4y ago
Render (https://render.com/ https://render.com/) is pretty nice.
hugg 4y ago
we're migrating to something called cloudflare pages, but i'm not involved in that
thibaut_barrere 4y ago
A couple of French-based options that I've been using as replacements: - https://www.clever-cloud.com https://www.clever-cloud.com - https://scalingo.com https://scalingo.com - https://www.qovery.com https://www.qovery.com

soperj 4y ago

Begs the question, what do people not like about Porter?

chimen 4y ago

Why companies move on to Heroku should be the question. All raised issues are there since Heroku.

rememberlenny 4y ago

Porter customer support is insanely good. If you are trying to migrate from Heroku to AWS, they make it incredibly easy and go out of their way to engineer alongside you. I was deeply impressed as a user and have nothing but positive things to say about their team.

craigkerstiens 4y ago

As someone that was there at Heroku after the acquisition, I don't think you can state that since the acquisition it's been consistently downhill. There isn't much debate that things have stagnated in the last 5-7 years which is a longer story on why probably. Some examples of innovation that happened and were launched after the acquisition: buildpacks (at the time of acquisition Heroku was still Ruby only), Heroku Postgres launched forks/followers/dataclips all after the acquisition, review apps came several years after. Salesforce may have had an eventual hand in it, but there was still a lot of innovation happening due to the folks there in the near to mid-term after the acquisition for several years. All that said, very excited for the new crop of players in the space. There are a number of companies trying to be a cheaper or more stable Heroku. Personally I'm excited about the ones that are taking their own unique approach. https://www.fly.io https://www.fly.io and https://www.railway.app https://www.railway.app are two that to me seem to bring their own perspective vs. just trying to recreate Heroku as a carbon copy clone. There are a number more in the jamstack space that have become staples such as Netlify and Vercel which are also doing great things.

booi 4y ago
https://render.com https://render.com is my favorite in this space
- bravura 4y ago
  Care to explain why you like these choices?
  - kjksf 4y ago
    I can only speak about render.com as I use for a mix of 10 static websites / Go webapps + cloudflare for DNS and caching proxy. It sounds prosaic but "it just works". Specifically compared to Digital Ocean Apps (which I used before render): * the dashboard UI is better designed and faster * the builds and deployments happen faster * similar price (to Digital Ocean, much cheaper than Heroku) * similar capabilities (attached disks, hosted Postgres, hosted redis) but the velocity seems better i.e. render.com seems to implement features faster than DO You would think that "well designed dashboard that displays instantly" would be a table stakes in an offering like that, but sadly it isn't.

jimmar 4y ago

I took a few minutes to scan Porter's website. They have a free tier. But when I read the getting started documentation, it says that I have to provision everything in my existing cloud infrastructure (e.g.; AWS). So I'm paying for the services in AWS. It seems I got lost in the story. Porter doesn't immediately feel like a replacement for Heroku.

jusrhee 4y ago
OP and Porter founder here. The article was meant to outline the most common technical limitations we see companies on Heroku bump up against as they outgrow Heroku. For individuals and teams running smaller workloads on Heroku where saving $ is a chief concern, Heroku is probably still a good option even though they’re declining in market share (this unprecedented recent outage aside). Porter is designed for companies that are maturing off Heroku for the technical reasons we mention or for those already looking to get the automation of Heroku in their own AWS/GCP cloud.
asenchi 4y ago
Make sure you are careful of their LICENSE if you choose to self-host. That repo is scary from a business perspective.

gkoberger 4y ago

I'm historically a huge Heroku defender. Been using it for a decade now, and my company still uses it despite being quite large and getting a lot of traffic. It's always been a great product and the early Salesforce days they DID ship a ton of new stuff and improve rapidly (despite a narrative that they didn't). Like, it got REALLY GOOD the years after acquisition. That being said... it's insane that we haven't been able to deploy for over two weeks and nobody there seems to care. So we're looking to move now, since it's clear Heroku has pretty much just given up at this point.

krallja 4y ago
> we haven't been able to deploy for over two weeks What is stopping you? Are you unable to type `git push heroku master`?
- tommoor 4y ago
  That's fine as a solo developer – but what about a team with CI pipelines etc?
  - VWWHFSfQ 4y ago
    can't you put that in the ci pipeline
    - scrose 4y ago
      I’m pretty confident if ‘putting it in the ci pipeline’ were a straight-forward option for most people, they probably wouldn’t be paying Heroku to manage review apps. I’ve used Heroku review apps for years and have also written and taken over different custom deployment pipelines. Review apps have a million different ways to be a giant time and money sink if not planned and implemented properly.
- gkoberger 4y ago
  I think you're underestimating the complexity of our codebase. Aside from the usual blockers, we have hundreds of repos (our Staging feature for Enterprise allows customers to deploy our code on their own cadence) all tied together using pipelines. We have our own custom release management software, which now doesn't work. Different repos have to go out at the same time so things don't break. Plus, we extensively use their review apps for code reviews, which we've lost access to. Lastly, not everyone has access to deploy directly to Heroku, so not everyone would be able to 'git push heroku main'. Could we fix all of this and get it working? Yeah. But we want to be focusing on building our product, which is why we pay Heroku a ton of money so we don't have to worry about this.

dml2135 4y ago

I'm at a small startup (about 8 engineers) and for the most part Heroku has been a great solution for a team of our size. However, the recent github incident was a major PITA for us, and has us seriously considering leaving the platform for the first time. We're still working on getting all of our automated deployments back up and running. In addition, we had just put in some work to start using the Review Apps feature, which now seems to be gone for the foreseeable future.

fomb 4y ago
Github actions are your friend friend: https://github.com/AkhileshNS/heroku-deploy https://github.com/AkhileshNS/heroku-deploy https://github.com/pmbanugo/heroku-review-app-actions https://github.com/pmbanugo/heroku-review-app-actions
- dml2135 4y ago
  This is awesome, thanks!

vmception 4y ago

Netlify do everything that I had used Heroku for, for free. In nearly one click too, instead of a dozen levers.

ajhurliman 4y ago
Do you only use lambdas for compute, or do they host servers now?
- vmception 4y ago
  They can have backends too, not just stuff rendered for the client You can also have those discreet compute instances as well, Netlify I think just recently introduced this, not sure how recent To my knowledge the only thing missing is a persistent storage solution, so you still need a database elsewhere but your api can read and write to it I personally dont do system design that way with my “web3” sites, as these are basically frontends to smart contracts, so I’m essentially using nearby nodes as compute instances and users pay to update the state of the compute instances and what the client therefore displays. Surprisingly economical for developers, I see that model attracting more devs very quickly as the stack and deployment is less complicated and cheaper and customers are already there.

intothemild 4y ago

Other than Dokku how many other PaaS solutions can be run on your own hardware?

c17r 4y ago
- piku https://github.com/piku/piku https://github.com/piku/piku - caprover https://caprover.com/ https://caprover.com/
jnettome 4y ago
we have been using caprover for 6 months, until we got the first overload and I had to find what exactly was happening during an outage. I love the solution but decided to "go by myself" after this. currently using a combination of docker swarm, traefik (via service labels) and portainer. totally in love, doing good for our needs (1 monolith, 12 microservices)

arberx 4y ago

One thing this article is missing is the arbitrary "400MB slug size" that heroku limits your application to if you want to deploy.

jacobsenscott 4y ago
FWIW that's a "soft limit". I've deployed slugs over that limit without trouble. Idk if they have a hard limit.
- henryaj 4y ago
  Hard limit is 500MB.
  - fomb 4y ago
    Which isn't hard. Support can increase it, but there are some proviso's such as increased deploy/boot time

davidkuennen 4y ago

I don't know if I'm seeing things, but Heroku seems to be more expensive than GCP. I get a 8 GB RAM Postgres instance for 200$ on Heroku. On GCP I get a 4CPU 26GB RAM Postgres instance for 280$ (Cloud SQL). What's the USP Heroku provides?

RedShift1 4y ago
They were early in the devops game, and provided an easy way to run things without having to deal with hardware, networks, scaling, operating systems, maintenance, etc... yourself. Nowadays I think they are no longer running on innovation, but on inertia.
nicoburns 4y ago
Point it at your got repo, click to choose which tech stack you’re running, write a 1 line yaml file to tell it what command to run to start up, and you already have a server running and a build pipeline that deploys when you push your code.

crcastle 4y ago

Small nit: I believe this is incorrect: > More recently, all Git-based deployments (which is to say, virtually all deployments) to Heroku were blocked and review apps were halted for all users as a result of a GitHub OAuth token leak. It should read "all GitHub-based deployments". You can still deploy with `git push heroku main`.

jusrhee 4y ago
Ah yup good catch - just updated the article
- nomilk 4y ago
  It caught my eye too, but for a different reason, this bit doesn't seem right: > which is to say, virtually all deployments My understanding is if deploying with `git push heroku main`, that application's GitHub repository was not viewable by hackers (but those apps deployed through 'Heroku GitHub Deploys' were). (please tell me if my understanding is incorrect). I think most Heroku users would deploy with `git push heroku main`, although that's purely hunch. Unrelated, but I'd add one more thing to the article, which is that Heroku docs aren't easy to give feedback on. I'd love for the docs to be on GitHub so shortcomings or inaccuracies can quickly be addressed. Currently, to point out a correction to the docs, you'd have to write a support ticket and 100% chance that support ticket isn't going beyond the person who received it, so nothing will get actioned.

justjake 4y ago

Warning: There are a couple people in this thread mentioning products that they're investors in. I won't call them out expressly, but I think people should be direct about their incentive alignments. It's disingenuous and borders on astroturfing IMO. Background: Founder of Railway.app here. There's a lot of these companies popping up that offer a "Heroku replacement", and once you dive in, you realize you have to pay $300/mo for a Kubecluster + $200/mo for the wrapped Kube service In our experience, people move off Heroku for a couple things: - Cost: Kinda self explanatory but Heroku pricing ramps hard - Flexibility: Heroku's not great for anything beyond stateful monoliths - Scalability: Notoriously Heroku's SLAs aren't that great In my mind, you don't replace Heroku with a minimum $500/mo Kubernetes cluster. Not only is this cost prohibitively expensive, but Kubernetes itself is a jet engine, and if you're not trained to use it correctly, you can risk catastrophic failure (on costs ballooning, on dataloss, etc) We're working hard at Railway to provide not just a Heroku replacement, but a next generation, composable infrastructure canvas. $5/mo, 30 seconds, and you're up and running. Demo:https://user-images.githubusercontent.com/5499880/165187948-8127b299-de74-49bb-81d4-c2e57ba66833.mp4 https://user-images.githubusercontent.com/5499880/165187948-... Would love any feedback and thoughts people have about Heroku, my thoughts above, and what we're building :) EDIT: Wow, just went from 11 -> 8 upvotes. I suppose they didn't like the astroturfing callout EDIT2: Render isn't who I'm talking about. They're great.

heartbreak 4y ago
They’ve been shilling their R•nder and P•rter services on every single thread about this Heroku incident, and it’s frankly a problem that HN moderators should address.
- ryanSrich 4y ago
  What should HN do? People are free to downvote it until it's dead. That seems a more democratic way to handle it.
  - justjake 4y ago
    I'm such a HN normie does HN even have a downvote button? I can't find it

rubyist5eva 4y ago

So this is more of a sales pitch for Porter - which is fine, I didn't know it existed and these are pretty compelling reasons to try it instead of Heroku. For me - the real selling point is that yes, Heroku is expensive - but the cost of a competent devops engineer is much, much, much higher. So unless it's literally impossible to do something with Heroku that needs a specialized skillset - the "devops as a service" part of Heroku is worth every penny in my opinion.

ajhurliman 4y ago
Agreed. We have a small number of high value users so we’re actually running on the smallest non-free tier (mostly to avoid cold starts) with almost zero head aches. It’s like $15 a month. The only frustration we’ve hit was mentioned in the article: static IP for integration with 3rd party services (VGS), but we used an add-on and I think we’re still in the free tier, or it’s a few dollars a month. The cost? We were actually spending more on GCP due to ham fisted provisioning, plus a $150k dev ops guy. So ya, I’m pretty happy about the cost.

paxys 4y ago

This is just an ad for their own service. Heroku is doing fine. They provide exactly what they advertise, and the service has been rock solid for over a decade. Sure it doesn't fit everyone's use cases, but it doesn't need to. Ultimately platform-as-a-service has always been a dead end. Companies either want more control over the infrastructure (so use plain VMs or Kubernetes), or want to forget about servers entirely (opting for services like Lambda and now edge computing). Everything in the middle (Elastic Beanstalk, App Engine, Heroku) has been stagnant for a long time now.

jacobsenscott 4y ago
Taking over two weeks to fix their GH integration is not "doing fine.". I've been a huge Heroku booster for years, but you can't ignore this. The GH breakage is a huge red flag.

jacobsenscott 4y ago

For all heroku's frustrations (and I agree with all of them in the article), it is still the only thing that "just works" for a standard monolith web app. Heroku is not cheap, but it is still cheaper than a couple full time employees + aws. It is really too bad they aren't innovating - they are just burning up their 10 year lead in the space. I wouldn't start a new project on Heroku - not because of the cost, but because I don't expect them to last another 10 years.

spicybright 4y ago
It's honestly strange how no one has eaten their lunch yet. Besides the extra services they now offer (which you aren't required for hosting a simple app), you could probably have a team bang out the basics in a few months. You could even just be a layer on top of AWS and probably make profit from not many users, as long as you're cheaper.
- mixedCase 4y ago
  > You could even just be a layer on top of AWS and probably make profit from not many users, as long as you're cheaper. Isn't Heroku exactly that?
- jacobsenscott 4y ago
  There are a lot of startups that have banged out the basics in a few months. The are getting mentioned all over in the threads under this article. But the basics aren't good enough. Heroku looks simple, but does a lot, and pretty much does it all right. Their documentation alone is amazing and it would take more than a few months to create docs of the same quality.
  - rsanheim 4y ago
    Yuuup. That is the genius of Heroku. They make almost everything very simple and the tougher stuff relatively easy to understand and configure. All of that took many many years of iteration and refinement, which is why the competitors are going to take awhile to get there.

mark_l_watson 4y ago

A good article, except for one big thing: this is not an end user, rather a direct competitor. Public articles that criticize competitors always run me the wrong way. I haven't used Heroku in a few years, but it has served (using the Hobby plan) as a really low cost way to host web apps. I have been reading through the comments on alternative providers, and even though I haven't used it, GCP's Cloud Deploy looks interesting also (a very long time ago, I used AppEngine a fair amount).

vmception 4y ago
Check Netlify Or Vercel Best scaling in their free plans Heroku hobby is a joke in comparison and hasnt been updated in a decade while all their addons have gotten less and less featured while costing more and more I host all my static assets on IPFS which practically nullifies the bandwidth limits of Netlify
- lambdaba 4y ago
  Could you elaborate on how you use IPFS? How fast is it? Do you use use a pinning service?
  - vmception 4y ago
    I upload static assets to IPFS and add those static links to my project instead of having an "assets" folder. IPFS is fast enough, I had these concerns too so I use a gateway for links as some of them cache files on their own CDN, for free. Yes, I use a pinning service, I was extremely hesitant as many of them are just SaaS hosting like Pinata is and charge by bandwidth used, but there are free ones like https://web3.storage https://web3.storage which for some reason is combining Filecoin and IPFS together for each upload and liveness/pinning. Ironically, Pinata also acts as a gateway and can then be used for free if you want. But there are other gateways. Cloudfront has one but what it decides to cache is strange.
krthr 4y ago
GCP Cloud Build + Cloud Run is a pretty good alternative for Heroku

digianarchist 4y ago

The year is 2022 and still no HTTP2 support. https://help.heroku.com/JAOCNZ25/does-heroku-have-plans-to-support-http-2 https://help.heroku.com/JAOCNZ25/does-heroku-have-plans-to-s...

multiplegeorges 4y ago
Or automated SSL for wildcard domains. > Supporting wildcard domains is something we'd like to do. Up until March 13, 2018, this was an upstream limitation from Let's Encrypt. ... 2018.
Beltalowda 4y ago
The changelog that's linked in there is depressing: only updates for languages/environments and an occasional minor update to the actual platform every few months.
jacobsenscott 4y ago
Yes, this is :(, but you should be serving all your static assets through a CDN, which will have http2. I can't think of a scenario where I would want to serve anything other than the dynamic content off the heroku server. I don't think http2 buys you much there. So you should be able to get away with at most 2 requests - one to the heroku server, and one to the cdn.
- mrkurt 4y ago
  The 2022 answer is actually to bake the CDN into your PaaS. Requiring a whole separate CDN for static assets is complication no one needs in their life.
- bpicolo 4y ago
  Need http/2 for grpc, which is certainly gaining popularity these days

intrepidsoldier 4y ago

In the last decade, a vast majority of DevTools companies have positioned themselves as “A better Heroku.” And yet, when people actually need a Heroku alternative, everything comes up short. This is a testament to a solid product Heroku built. It was definitely ahead of its time.

leetrout 4y ago
Bingo. Porter is one of the only ones that seems to offer feature parity across aws and gcp. Render.com looks cool but is also AWS only.
- anurag 4y ago
  (Render employee) We currently use AWS, GCP, and Cloudflare under the hood.
  - leetrout 4y ago
    Thanks for clarifying!

zoomzoom 4y ago

First off, I would strongly disagree with the idea that Heroku post-acquisition was stagnant. As well-said elsewhere, tons of product was shipped to industry-best standards: postgres/dataclips, Heroku Connect, buildpacks - as well as introduction of CI and preview apps. But it's clear that in the post-k8s and post infra-as-code world there's been a surge in new options for how teams manage their infra and DevOps toolchain. Of course, I’d agree with the idea that conflicts of interest should be disclaimed: I’m the co-founder of Coherence (https://www.withcoherence.com https://www.withcoherence.com). Seems the key issues raised by the article and comments here are: - costs and resource control constraints when using Heroku - fixed base costs of k8s when using self-hosted PaaS type systems - flexibility constraints of heroku-similar PaaS - the desire for teams to get more ownership than Heroku gives them when it comes to configurability, reliability & security Coherence disagrees that the next generation of PaaS should be a black-box - like Heroku, many next-gen PaaS are not hosted in your own cloud and re-implement the wheel when it comes to functionality like persistent storage, hosted databases, and application support services like redis. In the end, we believe that building on top of the major clouds (Google, AWS, Azure, CloudFlare, etc…) is the right choice. It’s also important to us that you're able to host customer data in mature systems that you control. This philosophy also allows Coherence to help you use managed services the large providers have built, like Cloud Run or App Runner, which at least partially mitigate the cost and complexity risks of k8s. Coherence is vertically integrated, composable, and opinionated, with a focus on developer experience. A defined workflow for production-quality full-stack web apps with dev and production built in alongside automated test environments, including CI/CD and cloud IDEs - all configured with one high-level YAML. We’re in a very early private beta on google cloud right now - if you’re interested, please check out our site above and let us know!

jeremyjh 4y ago
I was just looking at AWS App Runner today and just thinking about how miserable it would be to figure out how to use AWS's proprietary build pipeline tools to replicate what I have with Heroku - and I've been using AWS forever - I just hate their UX, their docs, really every thing they do except the actual infrastructure services. So I was just thinking someone should build a product that gives me the Heroku CLI, pipelines, add-ons, review apps etc but implements it all on top of standard AWS services. No runtime services at all, just orchestrating configuration and metadata, service provisioning etc. I know there are other companies in this space but I wish someone would just blatantly steal the Heroku UX and then sell me that.

tills13 4y ago

This seems like more of an ad / hit-piece than a discussion piece.

aantix 4y ago

I've heard good things about using Hatchbox to deploy to AWS. https://www.hatchbox.io/ https://www.hatchbox.io/

bradleybuda 4y ago

We're big Heroku customers; despite that, I largely agree with the points in this article and there's a voice in the back of my head asking every few months "is this worth it?" Whenever I research the new crop of Heroku clones (the one being hawked here, and others) the pitch is always "it's just like Heroku but you can run it in your own cloud". It's mind-boggling to me that none of the clones understands that I DON'T WANT TO RUN IT. Yes, I pay Heroku a premium because I like their software (the pipelines are great, dyno formations mostly Just Work) but what I'm really paying for is: * Never typing "ssh" * Never thinking about a full disk from a runaway log file * Never thinking about a load balancer or a certificate * Never waking up because a Postgres host has failed * etc, etc I have no interest in a "Heroku but you run it" PaaS but I'd pay though the nose for a "Heroku but it's actively developed" PaaS.

dontlaugh 4y ago
AWS’s ECS Fargate with Aurora Serverless and AWS CA certificates is close.
- ketzo 4y ago
  But even that stuff is gonna be more configuration than Heroku, right? I wanna spend literally zero hours a month on server / dev ops, if I can manage it. I will pay for it and accept the constraints of simplicity.
  - vlunkr 4y ago
    You can invest more time up front and automate it with an IAC tool and and CI tool to build/deploy. Probably nothing will be as hands off as Heroku, but there are lots of measures that can get you closer.
  - maccard 4y ago
    It is more configuration, yes, but it's far less than say ec2 and nomad. We run a very light production load on it and in 6 months I've had to intervene once (to bump our limits because we spiked slightly more than I expected us to)

wizwit999 4y ago

CDK +AWS is the way to go.

senjin 4y ago

Throwing this out there since it hasn't been mentioned yet but AWS's copilot tool (https://aws.github.io/copilot-cli/ https://aws.github.io/copilot-cli/) has been really nice for setting up an ECS cluster with a good build pipeline that we use in almost the same way as we use Heroku. It's definitely more mental overhead than Heroku but it offers more flexibility and tying directly to your AWS account. After setup it has been maintenance free for over a year but debugging parts of the build pipeline were painful at first. It recently started supporting AppRunner now that AppRunner can access VPCs so I'm looking forward to trying that at some point too.

chasd00 4y ago

Linode + dokku works pretty well for me

gwbas1c 4y ago

Awhile ago, my hobby Heroku site was shut down for vauge "violation of policy" reasons. Turns out it was an automated bot. When they brought it back up they kept breaking this and that... I kept having to re-open tickets. I'd really like a free alternative. (Or maybe I should figure out how to host from my NAS.)

caseydm 4y ago

Heroku is my tried and true. I've likely done thousands of deploys over the years and have never had one go bad or leave my app in a bad state. I have rolled back an app a few times due to bad configs and it was always quick and left me with a huge sigh of relief. That being said, they are very stagnant on features. A dyno has operated the same way, at the same size, at the same price for like 10 years now? The main features I would like to see are auto-scaling for all dynos and IPs that allow you to restrict database traffic. Render looks promising in that department and could likely get business from me if Heroku doesn't change in the next couple years.

rsanek 4y ago

If anything, this article solidified for me that the main reason remains cost only. Points (3) and (4) are probably not among the most important things for the vast majority of co's, even if they have 10s or 100s of employees.

ksajadi 4y ago

https://cloud66.com https://cloud66.com is the best alternative that runs on your own cloud and is native for Rails apps

antihero 4y ago

Have people that have grown out of heroku considered more dull but battle-tested solutions like CloudFoundry where you run a buildpack driven heroku-like abstraction on top of your own cloud provider (or whatever you want really)? Very widely used by enterprise, but also free and open-source.

ab-dm 4y ago

I built my first business on Heroku. Started 5 years ago and still, for the most part, it's going strong. There are some really strong downsides (especially after the last 2 weeks) but as a dev team of 1 (mostly) it's been a life saver, and that's enough to overrule all the other downsides. Having said that, we've suffered our own issues, with cost, latency (we're in AU) and scaling being the main ones. We're building a new product that we're betting very heavily on and it's all on AWS serverless. I wouldn't choose Heroku again. I think it was worth it 5 years ago, but as time goes on it becomes harder to justify.

pantulis 4y ago

key thing for me here is: why did Salesforce buy Heroku? How does Heroku fit into Salesforce strategy? When Heroku was incorporated by Salesforce there was no consumer facing capabilities in SF: everything was for business/sales users so Heroku was attractive to build public facing assets seamlessly connected to SF data (hence Heroku Connect). But then SF started incoporating other consumer facing featuers: Communities, Commerce (Demandware), heck they even have their own CMS now. One can predict they will push lowcode/citizen development beyond Force.com. Where does that leave Heroku? What does it bring to the table now for Salesforce?

rurban 4y ago

Losing GH automation did cost me exactly 20min. changing the alias from push github to push to both was trivial. I think Heroku did the right thing, but Travis still hasn't responded and changed anything, whilst the exploitation is still going on. If only I got get rid of Travis in my GitHub integrations. Deleting the app and actions and hooks hadn't changed anything. Travis has no access anymore, but PR's still block on the not existing Travis CI.

93po 4y ago

Heroku disabled my entire account with zero notice due to a fraudulant DMCA takedown notice, costing me tens of thousands of dollars. I struggled to be able to speak to anyone, and when I finally got a hold of someone, they treated me like a criminal despite making it clear I was the victim. Really shitty attitude, and the end result was getting my account reinstated but only with a "don't let this happen again, shithead" sort of treatment.

kasia66 4y ago

Usually, companies migrate from Heroku to cut the cost as they grow, but there are other reasons too: - The recent Heroku outages. - Lack of flexibility to adjust the available CPU, as Heroku offers six basic dyno types. As a result when the company grows they need to deal with overhears. - Heroku runs its servers on AWS, however, developers do not have access or control over the regions. This is an issue for the companies that deal with data requirements. - It does not offer support for crons/jobs. - Heroku default environment does not offer static IPs. (Private Spaces are only available on the Heroku Enterprise subscription.) Have a look at cloud66.com, it creates an environment like Heroku but on your servers on any cloud. This creates many benefits, including persistent storage and support for all available regions of your cloud provider of choice. But it also makes a big difference in availability: your application is not dependent on Cloud 66’s availability and won’t go down. Disclaimer:I work at Cloud 66.

cakoose 4y ago

Started something new last year and was considering Heroku, but their HA database offering uses async replication: https://devcenter.heroku.com/articles/heroku-postgres-ha#failover-conditions https://devcenter.heroku.com/articles/heroku-postgres-ha#fai... > Like followers, standbys are kept up to date asynchronously. This means that it is possible for data to be committed on the primary database but not yet on the standby. In order to minimize data loss we take two very important steps: > 1. We do not attempt the failover if the standby is more than 10 segments behind. This means the maximum possible loss is 160MB or 10 minutes, whichever is less. > 2. If any of the 10 segments were successfully archived through continuous protection, but not applied during the two minute confirmation period, we make sure they are applied before bringing the standby out of read-only mode. > Typically there is little loss to committed data. AWS RDS and GCP Cloud SQL do synchronous replication, so you are much less likely to lose data in common hardware failure scenarios. With those options available, a managed DB with async replication is a no-go. I don't understand how so many businesses would be ok with it. (I suspect most people don't even realize that Heroku HA is async.)