We are running production workloads on fly.io, but its getting increasingly unreliable. We had 1:30h of complete downtime just yesterday over a time span of 6 hours which is a LOT longer than the 35 minutes disclosed on the status page:
I undertand you are experiencing growing pains. But routing is a pretty substantial part of infrastructure. If I can’t trust the routing to be available then no other horizontal or vertical scaling, load balancing and fallback strategy will help.
I posted in another thread about not being able to deploy this morning. I’m not quite in production yet but have investor and client demos scheduled for today. I can wing it off of localhost in a screen share, but it’s definitely frustrating.
Ugh that is frustrating. I am coming back after a few weeks just trying to see if I can get this to work. What is the next best alternative? I’ve tried Gigalixer but that seems like ages ago
If you’re not bought into the rest of the Fly ecosystem (proxy, autoscale, multi-process, metrics, logs, litefs, consul, redis, postgres etc), then you could run the containers you deploy on Fly on AWS Lightsail Containers / AppRunner, Railway.app, Scaleway Serverless Containers, Vultr Managed Containers, Digital Ocean App Platform / Droplets, etc; Going multi-provider may be easy or hard depending on your setup.
I can understand that running a business like this can be invariable at times, but to have an incident be un-reported on their status page while also taking my services out for 24h+ is inexcusable
I’m not a paying customer because I don’t go past the loads neccesary, not because I don’t want to pay. Regardless, this is unacceptable. If there isn’t a good reason for this I don’t see any reason to continue with Fly.io
Thanks for the tips. When I said not working I was just talking about my app. I tried to deploy one on here a few weeks ago but got some out of memory issues, so I was just referring to that. So its just kinda frustrating to have Fly have their own issues
I might take a look around but I do like the approach here. Thanks again