☸️ The Deployment Looks Fine, and Nothing New Is Ever Scheduled
A namespace-level `ResourceQuota` set to cap total CPU, memory, or object counts can silently prevent new pods from being created once the namespace’s existing workloads collectively reach that limit – and because the Deployment or ReplicaSet controller itself reports no error (it correctly recognizes it needs to create a pod, it just can’t), the actual rejection reason is visible only in the specific pod-creation event, not anywhere higher-level status output shows by default.
🔎 The Problem
$ kubectl get deployment my-app NAME READY UP-TO-DATE AVAILABLE my-app 3/5 5 3 # 5 replicas desired, only 3 ready - the Deployment status alone gives # no indication of WHY the remaining 2 were never created. $ kubectl describe replicaset my-app-7d8f9c Events: Warning FailedCreate pods "my-app-7d8f9c-x9k2p" is forbidden: exceeded quota: compute-resources, requested: requests.cpu=500m, used: requests.cpu=4, limited: requests.cpu=4 # The namespace'"'"'s ResourceQuota caps total requested CPU at 4 cores - # existing pods already consume all 4, so every additional pod creation # attempt is rejected outright, with the rejection reason living only # in this specific ReplicaSet'"'"'s events, not in the Deployment'"'"'s own # summary status.
✅ Fix: Check the Quota, Then Decide Whether to Raise or Reduce
- `kubectl describe resourcequota -n
` shows exactly what’s currently being consumed against each defined limit in the namespace – the direct way to confirm whether a quota ceiling is actually the blocking factor before looking anywhere else. - If the workload genuinely needs more resources than the current quota allows, raising the `ResourceQuota` value (a decision usually made deliberately, in coordination with whoever manages overall cluster capacity, since quotas typically exist specifically to prevent one namespace from consuming unlimited shared cluster resources) resolves it directly.
- If the quota is set appropriately and the workload is the thing that should shrink instead, reducing the per-pod resource requests, or the desired replica count, brings the deployment back within the existing quota rather than requesting a limit increase that might not be warranted.
⚠️ Why This Is Easy to Miss During a Routine Scale-Up
- This tends to surface for the first time during what looks like an ordinary, low-risk scale-up (bumping replica count from 3 to 5) that’s worked identically many times before – the difference this time is that cumulative resource usage across the namespace has finally reached the quota ceiling, which isn’t something a routine scale-up would normally think to check first.
- Checking `kubectl describe replicaset` (not just the Deployment) is worth making a standing habit whenever a Deployment shows fewer ready replicas than desired with no other obvious explanation – the ReplicaSet’s own events are frequently where the actual, specific rejection reason lives.
A Deployment asking for more pods than a quota allows doesn’t fail loudly – it just quietly stops getting what it asked for, and the reason lives one layer deeper than where most people look first.

