Horizontal autoscaling for ate-apiserver #3706

Activity

1. Julian Gutierrez Oschmann (juli4n)

   added

   area/apiUser-facing API changesUser-facing API changes

   area/api-machinery

   on Nov 6, 2026
2. Julian Gutierrez Oschmann (juli4n)

   added the

   Feature

   issue type on Nov 6, 2026
3. laoj2 commented on Nov 6, 2026

   

   Luiz Oliveira (laoj2)

   on Nov 6, 2026

   Collaborator

   More actions

   Learnings from https://docs.cloud.google.com/kubernetes-engine/docs/how-to/design-substrate-infrastructure#size-system-infra-components:

   • Horizontal scaling stops improving throughput after a shared dependency (such as PostgreSQL CPU or connection limits) reaches capacity. Size your system infrastructure components and database together.
   • Each postgres replica opens up to poolMaxConns + 6 connections (owner 2, watch 3, readiness 1, in cmd/ateapi/internal/store/atepg/atepg.go). The sum across replicas must stay under max_connections, so the HPA maxReplicas should be floor((max_connections - reserved) / (poolMaxConns + 6)).
   • minReplicas >= 2, to keep today's availability
   • Candidate metrics: CPU utilization and RPC rate per pod.
4. Aditya Shantanu (aditya-shantanu)

   added

   kind/designDesign discussion, investigation, or research required before implementationDesign discussion, investigation, or research required before implementation

   on Nov 6, 2026