After this morning’s Cloudflare outage, I started wondering how other developers manage redundancy. Do you use backup servers or an automatic failover system so your application can switch to another environment when one provider goes down?
My app was completely unavailable for about 20 minutes because both Cloudflare and Supabase were affected, and unsurprisingly my client was not happy.
For those of you who rely on Supabase in the cloud, how do you architect high availability or disaster-recovery strategies? Any recommended setups, patterns, or best practices?
The user is seeking advice on implementing redundancy and failover strategies for their application, which was affected by a Cloudflare and Supabase outage. They are interested in understanding how other developers manage high availability and disaster recovery, including recommended setups and best practices.
Same question
I'm planning to self host auth and rest api but point it to supabase postgreSQL
then when an issue occurs, we could simply failover
In the dark ages, before we had cloud everything, this was solved by making everything redundant in different data centers. The layers upon layers of what you have to do is almost endless, including redundancy at each data center itself.
At each data center, we had twin fail-over routers, twin database servers, twin (or more) load balanced app and web servers. The DB servers would use streaming replication (at the time I used Slony-1 extension to Postgres) to create a backup which could be activated on failure of the primary. The firewalls would auto-fail over since they were mostly stateless, as were the web and app servers.
On top of that you need global redundancy on your DNS at multiple providers in multiple locations. You need to be able to load balance or route traffic to the different data centers.
We had nothing like Cloudflare to provide anycast routing and DDoS protection. You just had to hope and pray your upstream could handle it.
Running all of that is extremely expensive in both physical hardware and in people resources. Not everything can be fully automated.
So when you say backup and failover, you really are asking about every layer of your stack. At some point you just need to rely on the cloud vendors to do it for you. This is what I do now, and I take with it the risks of them failing from time to time.
That said, one thing I would really like from Supabase is a multi-region failover option using streaming replication, ideally using multiple clouds. I don't see how you get out from the need of a Cloudflare-like service for DDoS and bot protection and that cannot realistically be self-hosted, so you will always be at the mercy of someone.