2.50.0

New Features

  • It is now possible to attach a container to a Docker network other than the bridge or host network as well as to define network aliases for that container when using the Native scheduler. When specifying a user-defined network, it is a requirement that the network already exist.

Bug Fixes

  • Fixed an issue that would allow the Admin Console browser session token to be reused after logout.
  • Fixed an issue that could result in the replicated-ui container using increasingly high cpu.
  • All known CVEs with fixes have been patched in all images distributed by Replicated at the time of the release. For more information see this article.

2.49.1

Notes

  • This release is currently only available for the Kubernetes scheduler.

Bug Fixes

2.49.0

New Features

  • Most Replicated container exposed ports (excluding Replicated UI, Replicated Registry, and Graphite ports) can now be bound to the docker0 bridge network for single node Native Scheduler installations by passing the disable-replicated-host-networking flag to the install scripts.
  • When using the Native Scheduler, Replicated can now be installed in headless mode by passing the disable-replicated-ui flag to the install scripts.

Changes

  • Replicated Embedded Kubernetes version has been upgraded to 1.15.12.
  • It is now possible to restore a snapshot taken from a cluster running Replicated version 2.32.0+ onto the latest version of Replicated.
  • Restoring a snapshot will no longer override the snapshot destination settings from the snapshot, preserving the snapshot settings from the new installation.
  • The name of the Node onto which the bundle has been downloaded has been added to the messaging when running the replicatedctl support-bundle CLI command using Replicated Embedded Kubernetes.
  • A message has been added to the Support Bundle download page indicating that downloading the bundle may take several minutes.

Bug Fixes

  • Fixed an issue affecting the Swarm Scheduler that would cause restores via the CLI to fail with the error “Nodes failed to connect”.
  • Fixed an issue affecting the Swarm Scheduler that could cause Test Procs to hang resulting in timeouts.
  • Fixed an issue that would leave Replicated in a bad state if the network connection drops while uploading an airgap bundle, preventing the installation from proceeding.
  • Fixed an issue affecting the Native Scheduler that would prevent the Replicated Operator from connecting to Replicated when a more recent version of Replicated was previously installed.
  • The Replicated Embedded Kubernetes install scripts will now restart the Kubelet after Rook is ready, fixing a race condition that could cause the Kubelet to fail to detect the Rook Ceph FlexVolume plugin.
  • All known CVEs with fixes have been patched in all images distributed by Replicated at the time of the release. For more information see this article.

2.48.1

Notes

  • This release is currently only available for the Swarm and Native schedulers.

Bug Fixes

  • Fixed an issue introduced in 2.48.0 that would cause Swarm preflight checks and test procedures to default to restart policy “on-failure” causing unexpected results.
  • All known CVEs with fixes have been patched in all images distributed by Replicated at the time of the release. For more information see this article.

2.48.0

New Features

  • The property container.disable_publish_all_ports has been added to the Native Scheduler spec to disable publishing all exposed container ports that are not explicitly bound by the container.ports property.
  • The property container.oom_kill_disable has been added to the Native Scheduler spec to disable the kernel OOM killer for the specified container.

Changes

  • The Replicated shared filesystem initContainer will now delete its own pod when the shared filesystem fails to mount.
  • An hourly task has been added to clean up Replicated utility containers that have existed for more than one hour.
  • Time spent pulling and pushing the Programmable Preflight Check image to the On-Prem Registry will no longer count towards the timeout on the Swarm Scheduler.
  • Weave Net has been upgraded to version 2.7.0.

Bug Fixes

  • Fixed an issue that would allow expired airgap licenses to be installed.
  • A lock has been added to prevent multiple instances of the Port Available Native Scheduler preflight check to run at the same time, which would cause the second to report a failure.
  • Fixed an issue that would cause control plane pods to fail with ImagePullBackoff on Replicated Embedded Kubernetes primary nodes added to an airgapped cluster with the node-join script.
  • Fixed an issue that would cause the Replicated Embedded Kubernetes installation script to fail silently when the Docker cgroups driver was preconfigured. The script will now continue without reconfiguring the cgroups driver.
  • Fixed an issue that could cause Swarm Programmable Test Procedures to fail when the app is first installed because the image did not exist yet in the On-Prem Registry.
  • Fixed an issue that could cause Swarm and Native Preflight Checks to fail when the application is first installed because the On-Prem Registry is restarting.
  • Fixed an issue that could cause restoring with the Replicated CLI to fail due to not waiting long enough for remote Operators to connect to the primary.
  • All known CVEs with fixes have been patched in all images distributed by Replicated at the time of the release. For more information see this article.

2.47.0

New Features

Changes

  • An initiative has been made to move to a more inclusive vocabulary. This includes moving away from the master/slave terminology in favor of the terms primary/secondary when appropriate.
  • All Debian based Docker images have been upgraded from Debian 9 “Stretch” to Debian 10 “Buster”.
  • All Alpine Linux based Docker images have been upgraded from version 3.11 to version 3.12.
  • Replicated Embedded Kubernetes will use systemd as the cgroup driver for Docker and Kubelet on new installs.
  • Certain errors encountered during the prune phase after starting the application will be logged as errors rather than propagated to the application status when using Replicated Embedded Kubernetes .
  • Increased the CPU and memory requests for the weave-net DaemonSet deployed with Replicated Embedded Kubernetes.
  • Image pull concurrency within Replicated daemon is now capped at 4 concurrent image pulls to improve stability.

Bug Fixes

  • Fixed an issue where Docker 17.03 and 17.06 would fail to install due to the removal of the dockerproject.org apt and yum repos.
  • Fixed an issue that would cause the Replicated Operator to fail to reconnect when the host certificates were changed.
  • Fixed an issue with the Swarm scheduler that caused Support Bundles to remain on the filesystem indefinitely. Support Bundles older than 1 hour should be deleted by Replicated.
  • Fixed an issue that would cause the Premkit container to crash when a private key was included in the PEM certificate file.

2.46.2

Notes

  • This release is currently only available for the Kubernetes scheduler.

Bug Fixes

  • Fixed an issue that could cause the Replicated container to fail to start when Replicated is upgraded from a previous version and Kubernetes is configured to use a custom service subnet.

2.46.1

Notes

  • This release is currently only available for the Swarm and Native schedulers.

Bug Fixes

  • Fixed an issue that would cause airgapped installations to fail on older versions of Docker including 1.13.1.

2.46.0

New Features

Changes

Bug Fixes

  • Fixed an issue that could cause a Replicated Embedded Kubernetes Control Plane node to come back as worker nodes when rejoining the cluster.
  • Fixed an issue on Native and Swarm airgapped installations that could cause the Support Bundle to fail on secondary nodes due to failures to push the Support Bundle image to the on-prem registry.
  • Fixed an issue on the Native Scheduler that could cause the Operator to disconnect and fail to reconnect after generating a support bundle.
  • Fixed an issue that could cause the Support Bundle docker.container-exec task to fail with the error “Exec command is already running”.
  • Fixed an issue that could cause Kubernetes API server certificates to be generated with incorrect SANs on secondary nodes after changing the API server load balancer and using a non-standard service CIDR.
  • Fixed an issue that would cause the Kubernetes Ceph dashboard to be inaccessible on servers with IPv6 disabled.
  • All known CVEs with fixes have been patched in all images distributed by Replicated at the time of the release. For more information see this article.

2.45.1

Notes

  • This release is currently only available for the Kubernetes scheduler.

Bug Fixes

  • Fixed an issue that would cause the Kubernetes API server certificates to be generated with incorrect SANs when installed using an alternate Kubernetes Service CIDR.