Hi all , I am trying to setup spiceDB
# spicedb
d
Hi all , I am trying to setup spiceDB using container image in Azure Container Apps backed by Postgress Managed service. I couldnt find enough documentation for configuration there . Is this feature production ready and provide similar resiliency as spicedb operator ?
v
I'm not familiar with Azure Container Apps, is this the equivalent of AWS Elastic Container Service? When you say "this feature", what are you referring to? SpiceDB can likely run in Azure Container Apps, but it won't be able to cluster, and so it will loose it's horizontal scalability properties.
If azure provides an API for service discovery, we could potentially make clustering work on for Azure
d
By this feature I meant spicedb container image without spicedb operator. Azure container apps is like a wrapper over shared kubernetes. We can deploy ingress and scaling rules. So it could scale based on properties we mention but we dont have access to tune other k8s features like setting up liveliness probes for grpc calls , setting up topologyContraints . Its more like Fargate in AWS.
y
fargate is just a provisioning mechanism - it doesn't describe how the container is run
i'd assume that it's like ECS
and i'd recommend against running spiceDB using it
there isn't a good way to make SpiceDB containers aware of each other for horizontal dispatch in such an environment, which means you're leaving a lot of performance on the table
d
I need to know more about this. I could not find enough documentation or resources on this topic.Using spicedb container images in Production without spicedb operator. Do you have any references ?
y
no - this is what i've learned from running it in ECS and having conversations about enabling dispatch with the authzed devs
azure might be better than ECS in the sense that they might provide an XDS service discovery endpoint
but in ECS the only way to make containers aware of each other is through a load balancer (bad) or using DNS (less bad but still not good enough to make dispatch work well)
v
You can totally deploy SpiceDB how you would deploy any stateless application. We don't have documentation around this use-case tho, but fundamentally what you want to set is all the spicedb flags that configure the datastore, and it should work tm
some folks are happy running it that way. I only wish those services provided an equivalent to the endpoint discovery in the Kube API