New Features

  • Support for Ubuntu 20.04 LTS (Focal Fossa) has been added to the Native and Swarm schedulers.
  • NodeLocal DNSCache can now be enabled, to work around a bug resulting in DNS lookup delays of 5 or more seconds, by passing the nodelocal-dnscache flag to the Kubernetes install scripts.


  • The node join scripts generated on the /cluster page of the Admin Console will set the HTTP proxy if configured.
  • The Native scheduler will no longer remove application containers with the –force flag.
  • The default version of etcd bundled with Replicated Embedded Kubernetes has been upgraded from 3.3 to 3.4.
  • The container.stop_timeout property will now be passed to the container run command via the –stop-timeout flag in addition to the container stop command –time flag.

Bug Fixes

  • Fixed an issue in the Native scheduler that could cause application containers to be removed abruptly without sending a termination signal.
  • Fixed an issue that caused the Support Bundle to fail to collect when the Support Bundle Collect Spec is too long.
  • Fixed an issue that caused image layers from previous versions of the application to remain on disk in the Replicated Embedded Registry database after upgrade.
  • The Kubernetes install scripts will use UTC rather than the system time when creating the registry storage bucket.
  • The Kubernetes install scripts will now fail if there is a failure to create the registry storage bucket.
  • Fixed an issue in the Native scheduler that in some cases caused the Premkit container to fail with error “open /var/lib/replicated/secrets/premkit.crt: is a directory”.