🤔 While I like fly.io, I think they are encouraging/pushing people to use Server-Side SQLite because their managed Postgres support is hot garbage. 🗑️
I say this as someone who is a member of three orgs that have had severe Postgres database issues with their managed service. Support was not even remotely helpful.
I was even forced to pay $29 to get support for one of my personal projects just to be told they could not help me.
@webology I have been using Render.com for several prod web apps with their managed Postgres and it has been great. For the DB they only have a 3 month long free tier but other then that it has also been great for person projects. To me render.com is closer to Heroku then Fly.io.
@kedare@carlton death, taxes, and aws permissions... if only there were an app to... (mostly kidding)
We have used RDS for years with clients. Using AWS for personal web nodes sparks no joy for me, so I defer to shinier, simpler services like Fly (no db plz) and DO (some db plz).
@webology@carlton have you considered using AWS Elastic Beanstalk ?
It provides you an Heroku like DX but in AWS and much cheaper (you only pays for the AWS resources you use)
@kedare@carlton I did when it first came out, but I honestly don't want to spend anything with the AWS console when I'm not being paid to. I use KMS for my personal projects and everything else is either on a big DO droplet running Docker off a managed DB or I have been playing with fly since the setup/deploy is the most promising SaaS out there IMO.
I was afraid to mix the two out of latency to the DB concerns, but I'll probably sidecar pg-bounder going forward and not fear that as much.
@webology@kedare@carlton we use PostgreSQL (and Redis) managed instances in Digital Ocean and they work pretty well, are stable, easy to configure (e.g. replica, pool, ...), are cheaper then AWS and they don't have strange name for PostgreSQL. 😉 #postgresql#digitalocean#redis#devops
@markwalker@paulox@kedare@carlton they didn't know. It's strange to see RAM usage so low and then it spike like that. To me it feels like some backup job or maintenance job having issues. We didn't have any blocking or major processing running when this spikes. Otherwise, it should fire off with our scheduled tasks run.
@carlton I can deal with the IAM, but I don't want to. Amazon doesn't need my money and I find their business practices in this space to be vile and I can't unsee it. So I am voting with my feet/hosting despite my modest $4 a month S3/KMS bill each month.
@webology Ooh, that’s surprisingly political. OK. I’m squinting and the whole thing looks like capitalism to me — i.e. I’m not seeing any option that’s not distasteful if you’re actually going to look — but we can chat about this when next we meet. I was just picking up on your point about permissions, but that was a red-herring, no problem, thanks!
@carlton All good. It's been a while since I had to configure an IAM for ECS. I think the age of ChatGPT, I suspect most answers can be quickly handled there. I have been all Docker for a while so as long as a hosting provider can run an image, or give me a big cheaper server, it's almost painless to move.
@apollo13@kedare@carlton I have not, but I should. They supported DjangoCon US last year and I only hear glowing, wonderful things about them. So I will move a few things over and maybe report back.
They seem fine for web nodes, but use anyone else but them to trust your database is where I landed.
I spoke with https://www.crunchydata.com earlier this week with a client and they were a much better fit for a managed database service. I have no affiliation with CD, but they know their shit.
So use fly.io for web nodes and crunchydata for Postgres.
I'm using DigitalOcean's managed database on several projects and they have their own issues, but I have never had to worry about support playing hot potato on my issue or forcing me to pay $29 just to tell me that they can't help me.
@webology there’s just enough tooling via the flyctl postgres cli commands that, even though they stress in the docs it’s not managed, it lulls you slowly into forgetting that you’re on your own. That is, until the cluster goes belly up when you’re on vacation and then you become acutely aware of how on your own you are.
@nafiul@josh is kicking the tires aka GLOWING about it so far.
I had another client and a few friends use it and like it.
I need to move Django Packages to it at some point. DO gives me a weird error about once a week that I suspect I will never see again after we switch. (I'm using 10% of our total disk space even so it's weird)
@nafiul Small world. Craig is who we met with this week. I have known him since he spoke at DjangoCon US and joined us for a PyCon US dinner way back when. Good people. They have a few employees out of LFK that I have yet to meet too because small world.
Add comment