The “registry outage” decision tree
- Confirm scope. Is the registry service down, or just this host’s network? Check from a second host.
- Identify the cause. Service down, network partition, DNS, TLS, credentials.
- Pause the deploy. Do not continue to create new replicas that cannot start.
- Check existing replicas. They are running with cached images; they keep serving until they crash or restart.
- Restore the registry. The fix depends on the cause: failover DNS, restore service, refresh credentials.
- Resume the deploy. Once the registry is reachable, complete the rollout.
Local cache as a fallback
For critical images, mirror to a local registry or pull-through
cache (e.g. registry:2 configured as a pull-through cache
against Docker Hub):
# /etc/docker/registry/config.yml
proxy:
remoteurl: https://registry-1.docker.io
Each host pulls from the local cache. If Docker Hub is down, the cache serves from its local copy.