Upgrades

symlinks do not work in CDSW 1.9.x and CDSW 1.10.x

"If /var/lib/cdsw is a symlink to another location, CDSW might fail to start. If this occurs, you will see various errors in the ds-vfs pod about MountVolume.Setup fail, or No Such Project" errors: mount.nfs: mounting 192.168.225.133:/var/lib/cdsw/current/projects failed, reason given by server: No such file or directory.

Workaround:

Avoid using a symlink for /var/lib/cdsw.

Cloudera Bug: DSE-22863

Upgrades supported from CDSW 1.7.x (and higher) to CDSW 1.9.x

Upgrades supported from CDSW 1.7.x (and higher) to CDSW 1.9.x Cloudera Data Science Workbench only supports upgrades to version 1.9.0 from version 1.7.x and 1.8.x. If you are using an earlier version, you must first upgrade to version 1.7.x or 1.8.x, and then upgrade to version 1.9.0.

Issue with Upgrading CDSW in an airgapped environment

There is an issue with upgrading CDSW in an airgapped environment where docker service has no access to the k8s docker repository or has no proxy set. .

Workaround:

Upgrade to CDSW 1.9.1.

Alternatively, if you choose to continue using CDSW 1.9.0, you can complete the following steps. The following steps must be executed after deploying the CDSW parcel and DOCKER role is started but the MASTER, APPLICATION, and WORKER roles are not yet started. In case of an RPM install, this needs to be done after starting the dockerd service but before starting the CDSW master node.

Pull the missing etcd image to local docker repo in a node that has access to the k8s.gcr.io docker repository.
```
# docker pull k8s.gcr.io/etcd:3.3.15-0
```

Save the missing etcd image.

# docker save k8s.gcr.io/etcd:3.3.15-0 -o etcd.tar

Copy tarball to the CDSW master node.

# scp etcd.tar <user>@<node>:<location>/etcd.tar

SSH in to the CDSW master node and load the image to the local docker repo.
```
# docker load -i <location>/etcd.tar
```

Now you can start the rest of the roles for parcel installs and bring up cdsw cluster in case of an rpm install.

Cloudera Bug: DSE-13956