toomanyrequests: You have reached your pull rate limit

Symptom

During the bootstrap, one of the Bitnami / Grafana / Loki chart pulls fails with:

Error response from daemon: toomanyrequests: You have reached your pull rate limit.
You may increase the limit by authenticating and upgrading:
  https://www.docker.com/increase-rate-limit

Pods stay in ImagePullBackOff even though the image name is correct.

Root Cause

Docker Hub rate-limits anonymous image pulls to 100/6 hours per source IP. On a shared cloud provider your IP can burn through that in a single bootstrap if you're the second or third install on the same NAT egress.

The Bitnami charts, Grafana, Loki and Prometheus images all live on Docker Hub, so the whole cluster is affected.

Fix

1. Authenticate to Docker Hub (raises the limit to 200/6 hours)

Create a free Docker Hub account, then on the server:

docker login
# Login with your Docker Hub username + a Personal Access Token

For k3s / containerd (which is what the pods actually use), also drop the credentials into containerd's registry config:

sudo mkdir -p /etc/rancher/k3s
sudo tee /etc/rancher/k3s/registries.yaml > /dev/null <<'YAML'
configs:
  "docker.io":
    auth:
      username: <your-docker-hub-user>
      password: <your-docker-hub-personal-access-token>
YAML

sudo systemctl restart k3s

2. Mirror through GHCR (best for real production)

Every Bitnami image is mirrored to ghcr.io/bitnami/*. In Helm values, swap the repository — for example for MinIO:

image:
  repository: ghcr.io/bitnami/minio
console:
  image:
    repository: ghcr.io/bitnami/minio-object-browser

The stock bootstrap already uses bitnamilegacy/* mirrors for MinIO; you can apply the same pattern to Postgres, Mongo and Redis if you keep hitting the limit.

3. Retry the failed pulls

kubectl -n <namespace> rollout restart deployment/<name>
# or force the containerd cache to re-fetch:
sudo k3s crictl pull docker.io/bitnami/postgresql:16

Verification

kubectl get pods -A | grep -v Running | grep -v Completed

Empty output (except real errors you already knew about) means the pulls succeeded.