Longhorn v1.0.1-rc2 release

Highlights:

1. NFS on Longhorn for RWX Access Mode (experimental).
2. Improved Air Gap Environment Support. There are a few issues related to the enhancement.
   1. Remove the air gap image name length limitation.
   2. Change the default image pull policy from Always to IfNotPresent
   3. Aggregate the necessary Docker images
   4. HTTP/HTTPS_PROXY support for S3 backupstore
3. Support PriorityClass for Longhorn deployment
4. Make Longhorn Helm Chart directly available through Helm repo
5. Virtual Hosted Style S3 bucket support

Upgrade:

Live upgrade from v1.0.0 to v1.0.1 is supported.

Note: upgrade from/to release candidate is not supported.

Longhorn v1.0.1-rc1 release

Highlights:

1. NFS on Longhorn for RWX Access Mode (experimental).
2. Improved Air Gap Environment Support. There are a few issues related to the enhancement.
   1. Remove the air gap image name length limitation.
   2. Change the default image pull policy from Always to IfNotPresent
   3. Aggregate the necessary Docker images
   4. HTTP/HTTPS_PROXY support for S3 backupstore
3. Support PriorityClass for Longhorn deployment
4. Make Longhorn Helm Chart directly available through Helm repo

Upgrade:

Live upgrade from v1.0.0 to v1.0.1 is supported.

Upgrade from/to Release Candidate is not supported.

Longhorn v1.0.0 Release

Longhorn v1.0.0 is Longhorn's first General Available (GA) Release.

Highlights:

1. Automatically detach the volume for Deployment when the node is down. Currently
   Kubernetes has an issue that it will not automatically detach the volume from the pod if the node is down. We are planning to enable Longhorn to detach the volume from a Deployment pod if it's stuck on a down node by removing the VolumeAttachment automatically. In addition, we're planning to implement something similar to node fencing in Longhorn v1.1.0 to address a similar issue with StatefulSet. See here for details.
2. Support custom CA for Backupstore. This feature will allow the user to bring in their own CA certificate to Longhorn. This is useful for setting up a Backupstore in an air gap installation. See here for details.
3. Automatically disable Longhorn node scheduling when cordoned. Currently, Longhorn will continue to schedule more replicas to a cordoned node. This behavior may not be consistent with what users would expect. This feature will give users an option to define the desired Longhorn behavior, including disabling node scheduling automatically when a node is cordoned. See here for details.

Release Note:

1. The following settings' default value has been tweaked. Even though the change will only apply to the newly installed system, but we recommend following the new default setting value if possible. See the document for more details:
   1. The default replica anti-affinity setting has been changed from soft to hard. See here for details.
   2. The default GuaranteedEngineCPU has been changed from 0 to 0.25. See here for details.
      1. Notice the change of this setting will result in recreating all the instance managers on the node, so all the attached volumes will be detached and interrupt the data flow. We highly recommend the users to perform this change after making sure all the volumes are detached, e.g. during the upgrade/maintenance window.
2. The driver name io.rancher.longhorn used by pre-v0.7.0 release is no longer supported. You can follow the steps here to convert your old PVs before upgrading to v1.0.0

Upgrade:

The only supported upgrade path is offline upgrade from v0.8.1 to v1.0.0. See here for details.

Longhorn v1.0.0-rc3 release

Upgrade from/to a release candidate is not supported.

Highlights:

1. Automatically detach the volume for Deployment when the node is down. Currently
   Kubernetes has an issue that it will not automatically detach the volume from the pod if the node is down. We are planning to enable Longhorn to detach the volume from a Deployment pod if it's stuck on a down node by removing the VolumeAttachment automatically. In addition, we're planning to implement something similar to node fencing in Longhorn v1.1.0 to address a similar issue with StatefulSet.
2. Support custom CA for Backupstore. This feature will allow the user to bring in their own CA certificate to Longhorn. This is useful for setting up a Backupstore in an air gap installation.
3. Automatically disable Longhorn node scheduling when cordoned. Currently Longhorn will continue to schedule more replicas to a cordoned node. This behavior may not be consistent with what users would expect. This feature will give users an option to define the desired Longhorn behavior, including disabling node scheduling automatically when a node is cordoned.

Release Note:

1. The following settings' default value has been tweaked. Even though the change will only apply to the newly installed system, but we recommend following the new default setting value if possible. See the document for more details:
   1. The default replica anti-affinity setting has been changed from soft to hard.
   2. The default GuaranteedEngineCPU has been changed from 0 to 0.25.
      1. Notice the change of this setting will result in recreating all the instance managers on the node, so all the attached volumes will be detached and interrupt the data flow. We highly recommend the users to perform this change after making sure all the volumes are detached, e.g. during the upgrade/maintenance window.
2. The driver name io.rancher.longhorn used by pre-v0.7.0 release is no longer supported. You can follow the steps here to convert your old PVs.

Upgrade:

The only supported upgrade path is offline upgrade from v0.8.1 to v1.0.0. See the document for more details.

Longhorn v1.0.0-rc2 release

Upgrade from/to a release candidate is not supported.

Highlights:

1. Automatically detach the volume for Deployment when the node is down. Currently
   Kubernetes has an issue that it will not automatically detach the volume from the pod if the node is down. We are planning to enable Longhorn to detach the volume from a Deployment pod if it's stuck on a down node by removing the VolumeAttachment automatically. In addition, we're planning to implement something similar to node fencing in Longhorn v1.1.0 to address a similar issue with StatefulSet.
2. Support custom CA for Backupstore. This feature will allow the user to bring in their own CA certificate to Longhorn. This is useful for setting up a Backupstore in an air gap installation.
3. Automatically disable Longhorn node scheduling when cordoned. Currently Longhorn will continue to schedule more replicas to a cordoned node. This behavior may not be consistent with what users would expect. This feature will give users an option to define the desired Longhorn behavior, including disabling node scheduling automatically when a node is cordoned.

Release Note:

1. The following settings' default value has been tweaked. Even though the change will only apply to the newly installed system, but we recommend following the new default setting value if possible. See the document for more details:
   1. The default replica anti-affinity setting has been changed from soft to hard.
   2. The default GuaranteedEngineCPU has been changed from 0 to 0.25.
      1. Notice the change of this setting will result in recreating all the instance managers on the node, so all the attached volumes will be detached and interrupt the data flow. We highly recommend the users to perform this change after making sure all the volumes are detached, e.g. during the upgrade/maintenance window.
2. The driver name io.rancher.longhorn used by pre-v0.7.0 release is no longer supported. You can follow the steps here to convert your old PVs.

Upgrade:

The only supported upgrade path is offline upgrade from v0.8.1 to v1.0.0. See the document for more details.

Longhorn v1.0.0-rc1 Release

Upgrade from/to a release candidate is not supported.

Highlights:

1. Automatically detach the volume for Deployment when the node is down.Currently
   Kubernetes has an issue that it will not automatically detach the volume from the pod if the node is down. We are planning to enable Longhorn to detach the volume from a Deployment pod if it's stuck on a down node by removing the VolumeAttachment automatically. In addition, we're planning to implement something similar to node fencing in Longhorn v1.1.0 to address a similar issue with StatefulSet.
2. Support custom CA for Backupstore. This feature will allow the user to bring in their own CA certificate to Longhorn. This is useful for setting up a Backupstore in an air gap installation.
3. Automatically disable Longhorn node scheduling when cordoned. Currently Longhorn will continue to schedule more replicas to a cordoned node. This behavior may not be consistent with what users would expect. This feature will give users an option to define the desired Longhorn behavior, including disable node scheduling automatically when a node is cordoned.

Release Note:

1. The default replica anti-affinity setting has been changed from soft to hard. It will only apply to the newly installed system.
2. The driver name io.rancher.longhorn used by pre-v0.7.0 release is no longer supported. You can follow the steps here to convert your old PVs.

Longhorn v0.8.1 release

Highlights:

1. New Longhorn website. Please visit https://longhorn.io/docs for latest documentations.
2. Default node/disk configuration. This enhancement allows the user to customize the default disks and node configurations in Longhorn for newly added nodes using Kubernetes label and annotation, instead of using Longhorn API or UI. See here for details.
3. Air Gap installation support. We've added Registry Secret support and now Longhorn supports installation in Air Gap using the Helm Chart. See here for details.
4. Volume Expansion Rollback. With this feature, Volume Expansion is no longer an experimental feature.

The minimal Kubernetes version supported is v1.14.0.

Important upgrade notes:

1. Live upgrade is supported from v0.8.0 to v0.8.1. You can follow the instruction here to upgrade from v0.8.0. To perform the non-disruptive upgrade for Longhorn engine from v0.8.0 to v0.8.1, follow the instruction here after you upgraded the Longhorn manager.
2. Longhorn v0.8.1 supports upgrading from v0.8.0 only. Please upgrade to v0.8.0 first before upgrading to v0.8.1.
3. Driver name migration. During the upgrade from v0.6.2 to v0.7.0, we've changed our Kubernetes driver name from io.rancher.longhorn to driver.longhorn.io. In v0.7.0, we've deployed a compatible CSI plugin to accommodate the old io.rancher.longhorn driver provisioned volume. We're going to remove the compatible CSI plugin in the GA release. So we highly recommended the users that still using PVs provisioned by io.rancher.longhorn to convert the PVs to the new driver name. You can follow the steps here to convert your old PVs.

Longhorn v0.8.1-rc3 release

Upgrade from/to a release candidate is not supported.

The minimal Kubernetes version supported is v1.14.0.

Highlights:

1. Default node/disk configuration. This enhancement allows the user to customize the default disks and node configurations in Longhorn for newly added nodes using Kubernetes label and annotation, instead of using Longhorn API or UI. See here for details.
2. Air Gap installation support. We've added Registry Secret support and now Longhorn supports installation in Air Gap using the Helm Chart. See here for details.
3. Volume Expansion Rollback. With this feature, Volume Expansion will no longer an experimental feature.

Longhorn v0.8.1-rc2 release

Upgrade from/to a release candidate is not supported.

The minimal Kubernetes version supported is v1.14.0.

Highlights:

1. Default node/disk configuration. This enhancement allows the user to customize the default disks and node configurations in Longhorn for newly added nodes using Kubernetes label and annotation, instead of using Longhorn API or UI. See here for details.
2. Air Gap installation support. We've added Registry Secret support and now Longhorn supports installation in Air Gap using the Helm Chart.
3. Volume Expansion Rollback. With this feature, Volume Expansion will no longer an experimental feature.

Longhorn v0.8.1-rc1 release

Upgrade from/to a release candidate is not supported.

The minimal Kubernetes version supported is v1.14.0.

Highlights:

1. Default node/disk configuration. This enhancement allows the user to customize the default disks and node configurations in Longhorn for newly added nodes using Kubernetes label and annotation, instead of using Longhorn API or UI. See here for details.
2. Air Gap installation support. We've added Registry Secret support and now Longhorn supports installation in Air Gap using the Helm Chart.
3. Volume Expansion Rollback. With this feature, Volume Expansion will no longer an experimental feature.