Appwrite Horizontal Scaling

Hi!!
I’m planning to scale Appwrite horizontally and I wanted to ask if there are any updates or best practices in 2026 for doing this in multi-node or multi-region setups.

I’ve been considering using Docker Swarm to handle the replication of stateless containers (API, workers, realtime), but I understand there are some limitations like the appwrite-executor container that keeps the function state in memory, so replicating it can cause issues.

My idea would be to have a distributed database (MongoDB Atlas or Azure CosmosDB) so all nodes have access to the same data, and a centralized S3-compatible storage so all nodes share the same files.

I’d love to get advice on the best way to scale Appwrite horizontally today, how to handle Docker Swarm and the executor to avoid issues with functions, or any other robust alternatives you could recommend.
Thanks everyone!

Hey, I am curious, why do you need horizontal scaling?

Also, FYI appwrite does not use mongodb nor cosmosdb.

I want to scale Appwrite horizontally across multiple nodes and regions to handle more traffic, reduce latency by connecting users to the closest server, and improve availability if a node goes down.

True. But i think that Rds mysql (Or rds for mariaDb) could work

I see. I just asked cause that's usually only needed by big projects usually. I can't help much myself.

cc @Aditya Oberai Hopefully can help

Thanks!

I've not done it myself but inevitably there's going to come a point where I need to. I came across this guide. Not sure how relevant it is these days but it might give you some ideas? Was planning on having a go myself at some point: https://book.appread.io/development

Currently i have scaled multiple containers from appwrite (worker functions, executors, ...)

but more recently due to some others issues, i also tried to scale the API, RealTime and Console.
All 3 have been scaled successfuly.

Also i think all container could be scaled (except where there is storage like maria and redis containers) by simply adding more container with the same configuration.

here is my API service on my k8s deployment which currently have 2 pods/container running

and sorry for the french in the screenshot

also no my setup is not multi region like appwrite cloud have. but more for redundency and HA for a single region

Thanks a lot for the detailed reply, really appreciate you taking the time 🙏

I’m currently evaluating how to deploy Appwrite with high availability in a single region.

I’m quite interested in Kubernetes, especially after reading your experience, although I’m still a bit hesitant since Appwrite isn’t officially supported on K8s.

One of my main questions is around scaling:
– How do you replicate the workers across pods?
– And how do you handle Appwrite Functions code so that all worker pods can execute them consistently?

From your experience, did this part require any special setup or workarounds?

Thanks again for sharing your setup — it’s really helpful to hear real-world experiences like this.

– How do you replicate the workers across pods?

In my k8s deployment each worker have it's deployment and is "one worker = one pod".
If you meant replicate across nodes, then i don't really choose where they go but you could use the DaemonSet in k8s to set something like "X replicas with max one replica per node"

– And how do you handle Appwrite Functions code so that all worker pods can execute them consistently?

My functions are setup like this :

• Storage volumes are now RWX PVC
• multiple worker functions check for jobs, when one is present, one worker will take the job
• for the executor, i'm using the RWX PVC + original host binds like /tmp.

From your experience, did this part require any special setup or workarounds?

The scaling up didn't require any changes, the only thing appwrite team told me is that if you have multiple executor, when running sites, your site might run on multiple executors because it will load balance across them.

The big workarounds where for k8s (e.g. rwx for storage, dind for executor)

If you want i have a chart here : https://gitlab.ibaraki.app/helm/appwrite (not official, use at your own risk)

@Ibaraki Douji Huge thanks for you detailed explanation

Thanks so much for the explanation. Is so helpful. I will try use it