Engprod 10109 optimize clowder resource configs

[ffeghali]

Right-size hard-coded CPU/memory requests and limits for ephemeral-only containers deployed by Clowder. These containers only run when web.mode: local, featureFlags.mode: local, or reverseProxy.mode: ephemeral — so changes do not affect production or staging.

Resource values were determined using recommend_resources_v6.py, which queries the same Prometheus metrics behind the Grafana Container Resource Rightsizing dashboard. Data covers a 30-day window across 355 ephemeral namespaces, with the ADR-046 formula applied:

• CPU request = P95 usage * 1.2 / 2
• CPU limit = max usage * 1.2
• Memory request = P95 usage * 1.2
• Memory limit = max usage * 1.2

Changes Applied

Container	CPU Req	CPU Limit	Mem Req	Mem Limit
keycloak-db	50m → 23m	100m → 57m	100Mi → 58Mi	200Mi → 136Mi
mbop	50m → 1m	100m → 3m	100Mi → 20Mi	200Mi → 55Mi
mocktitlements	50m → 1m	100m → 1m	100Mi → 12Mi	200Mi → 23Mi
featureflags-db	50m → 16m	100m → 42m	100Mi → 59Mi	200Mi → 137Mi
featureflags-edge	100m → 7m	200m → 16m	200Mi → 24Mi	400Mi → 72Mi

Not included (pending discussion)

MinIO currently has zero resource constraints and is not ephemeral-only (objectStore.mode: minio can be set in any environment). Adding limits requires team discussion.

🤖 Co-written with Claude Code

[ffeghali]

/test-e2e

[ffeghali]

/test-e2e