We do, although we're in the middle of moving our entire Heroku Postgres spend over to Crunchy Data [1].
We were getting close to one of the big jumps on the standard pricing of Heroku Postgres, and we would have had to basically double our monthly cost to lift the max data we could store from 1.5TB to 2.0TB. On Crunchy Data, that additional disk space will be like 1% more rather than 100% more.
While investigating Crunchy, I ran some benchmarks, and I found Crunchy Bridge Postgres to be running 3X faster than Heroku Postgres.
Heroku seems to be working on some interesting new things, but I feel burned by the subpar performance and lack of basically any new features over many years. I don't know if the new Aurora-based database will be faster than Crunchy, but the benchmarks they're talking about sound like they're finally about to catch them. But we also have better features on Crunchy, too, such as logical replication. Logical replication is still not available on Heroku.
The experience for deploying apps and having add-ons is still pretty easy, but we'll see how that improves. HTTP2 support is still in beta.
My experience with going from Heroku Postgres to Crunchy Data (specific Crunchy Bridge) has been really good. Their product has been absolutely rock solid but what really made the difference was their support. They provided a huge amount of pre-sales support while I planned the move (and even suggested mitigations for the problems I was having with Heroku Postgres to make moving less urgent). Post-sales support has been just as good, though mostly I don’t even have to think about the database hosting anymore.
I also moved my app hosting to NorthFlank from Heroku and have been really happy with that as well. It’s got the features I always wanted on Heroku (simple things like grouping different types of instances together into projects really helps) plus again excellent responsive support.
Our experience of moving from Heroku to CrunchyBridge has been very similar - excellent help with the migration including jumping on a call with us during the switchover to resolve a broken index.
Would strongly recommend them to anyone looking to move off Heroku.
I was a bit concerned about the cut-over from the old database on Heroku, really wanted to minimise downtime. So they helped me produce a step by step plan, test as much of it as possible, then had an engineer join me on Zoom while I made the switchover. They were even able to accomodate doing it in the early morning in my timezone to minimise the impact. Ended up with maybe 5 mins of downtime which I was very happy with.
I'm working in a new startup, and I tried several "easy" solutions: AWS Lightsail, Heroku, Crunchy.
I settled up on AWS ECS :)
My main issue with Heroku was that they have not changed anything in _years_. No support for gRPC, no IPv6, and simple VPC peering costs $1200 a month.
Yeah, the lack of HTTP/2 support has been a long-standing issue with Heroku.
They just shipped HTTP/2 terminated at their router [0], and have it on their roadmap [1] to support HTTP/2 all the way through. But it seems like it's at minimum a few months off.
(As for VPC peering: the moment you need that, it sorta feels like Heroku is no longer the right place to be, even ignoring the costs.)
Update to this: we've switched over our staging database, and the call to do that couldn't have been. more productive or more pleasant.
I got to talk to someone who was intelligent about Postgres, who answered various questions I had, who offered a few pieces of insight that I wouldn't have thought, etc.
Compared to every single support interaction I've ever had with Heroku for 10 years, and this was light years more friendly, informative, and productive.
Despite interesting competition, my feeling is that the Heroku of 2024 remains... Heroku.
I feel this way even though -- depending on how you segment -- the list of "interesting" competitors is quite long at this point: Render, Railway, Northflank, Fly.io, Vercel, DO App Platform, etc.
I revisit heroku alts every ~6months and I am shocked how not ergonomic they still are, I switched to DO VPS + Ansible Container & Github Actions for any project that doesn't need infinite scale after Salesforce paused heroku development but I'd go back to literally any heroku clone.
It's crazy how the ergonomic still just aren't there.
Yes, completely agree; I'm equally surprised by the poor DXes.
(And: bugs. I'm also surprised by the kinds of issues I run into on some of those sites in my list -- problems that, even if not show-stopping, feel like revealing indicators of quality.)
> Yes, completely agree; I'm equally surprised by the poor DXes.
Any specifics/examples? I find it hard to imagine those "big name" companies/platforms you just mentioned don't have entire teams dedicated to hyper-optimizing experience.
Can you elaborate a bit more why render is good? we are on heroku and I have evaluated alternatives every 6 months since heroku/github outage 2 years ago [1]. But I don't see how render is better. 2 years ago render postgres did not have PITR. now they have build it, but Render's postgres offering is even more expensive than heroku, and queries run a bit slower on similar spec machines based on my test. I also don't like render charges per seat in addition to infra cost.
10yo+ B2B SaaS company we're still on Heroku. I think the value prop is particularly good for B2B SaaS and probably less so for consumer products. Our margin per customer is so much higher than the infra cost it just never makes sense to spend money on devops instead of building features.
That said it does feel a bit like a ghost town, I'm always happy to hear when someone is doing something over there.
A few years ago I was considering Heroku for something new. But then I learned that Heroku Postgres's HA offering used async replication, meaning you could lose minutes of writes in the event that the primary instance failed. That was a dealbreaker.
That was very surprising to me. Most businesses that are willing to pay 2x for an HA database are probably NOT likely to be ok with that kind of data loss risk.
(AWS and GCP's HA database offerings use synchronous replication.)
Noticed this too. The master failover is marketed like a strict upgrade, and the "async" part is only in the fine print. Many would actually prefer the downtime over losing data. A user who's experienced with DBs should think to check on this, but still.
I’ll just share our experience with Hetzner from earlier this week to spare everyone the learning experience. Their prices are indeed incredible. However, we spun up a VPS in their Hillsboro, OR data center only to find out that our IP address was blocked by Cloud Flare, so there was no way for us to connect to our error logging or transactional email providers. We also found this thread indicating that their entire IP range for that DC is widely blocked[0]. So, not really acceptable for professional use, IMO, unless you just need compute.
So in that case both Mailgun and Sentry were blocking our IP. When we googled the error message, it appeared to be a Cloud Flare message of some kind. But anyway, wouldn’t touch them with a ten foot pole for anything web related.
probably those who don't want to set uptime alerts, fine-tune configs, set up backups and restores (which are essential because sooner rather later someone always deletes a few rows/tables) and want to focus on business
It's really easy to be SOC2 compliant for a small SaaS on heroku. We need to grow in customers and Dev resources to pull it off on Raw AWS. Am looking for options though because heroku is increasing their prices.
me. tried to move to render but it's been a headache for some key things that i need. my heroku setup is dialed in so it makes it a no brainer versus the time i've wasted trying to get render to fit my use case. right now i'm using both for two different services, and will consider moving off both once i get enough customers.
Nothing sophisticated. My woes might be because I'm bad at devops.
But I spent several hours fighting with a DNS change, trying to host my marketing website as Cloudflare pages site from my root domain (with DNS managed by Cloudflare), and then wildcard subdomains routed to a Render server. I couldn't get it to work no matter what configs I tried. My root domain marketing site is proxied through Cloudflare and I was trying to get the wildcard subdomains as DNS-only, and I suspect this was the problem but idk. In other words, the Cloudflare pages marketing site is https://bookhead.net and I wanted my customer's subdomains to route to the Render server like https://forlornbooks.bookhead.net/ (I still get the error since I haven't finished my migration to Heroku). The subdomains worked with no problem until I tried to setup a separate marketing site at the root domain.
Also, I had a hard time setting up SSH with a containerized server. It was a weird DX that was a bit confusing to document so I can remember later. Can only imagine how confusing it might be if I ever have teammates. The Render CLI looks promising, though.
These are only the most recent issues. Seems like y'all improved the headaches I ran into the time I tried.
A VPS (hetzner, etc) + managed postgres DB (supabase / AWS / etc) or a local one might more more than enough these days.