Upgrade JuiceFS Client

Upgrade JuiceFS Client to the latest version to enjoy all kinds of improvements and fixes, read release notes for JuiceFS Community Edition or release notes for JuiceFS Cloud Service to learn more.

As a matter of fact, upgrading JuiceFS CSI Driver will bring upgrade to JuiceFS Client along the way, because every release includes the current latest mount pod image, but if you'd like to use the latest JuiceFS Client before CSI Driver release, or even before mount pod image release, refer to methods introduced in this chapter.

Upgrade container image for mount pod

Find the latest mount pod image in Docker Hub, and then overwrite mount pod image.

Pay attention that, with mount pod image overwritten, upgrading CSI Driver will no longer affect mount pod image.

Upgrade JuiceFS Client temporarily

tip

You are strongly encouraged to upgrade JuiceFS CSI Driver to v0.10 and later versions, the method demonstrated below are not recommended for production use.

If you're using Mount by process mode, or using CSI Driver prior to v0.10.0, and cannot easily upgrade to v0.10, you can choose to upgrade JuiceFS Client independently, inside the CSI Node Service pod.

This is only a temporary solution, if CSI Node Service pods are recreated, or new nodes are added to Kubernetes cluster, you'll need to run this script again.

1. Use this script to replace the juicefs binary in juicefs-csi-node pod with the new built one:

   #!/bin/bash
   
   KUBECTL=/path/to/kubectl
   JUICEFS_BIN=/path/to/new/juicefs
   
   $KUBECTL -n kube-system get pods | grep juicefs-csi-node | awk '{print $1}' | \
       xargs -L 1 -P 10 -I'{}' \
       $KUBECTL -n kube-system cp $JUICEFS_BIN '{}':/tmp/juicefs -c juicefs-plugin
   
   $KUBECTL -n kube-system get pods | grep juicefs-csi-node | awk '{print $1}' | \
       xargs -L 1 -P 10 -I'{}' \
       $KUBECTL -n kube-system exec -i '{}' -c juicefs-plugin -- \
       chmod a+x /tmp/juicefs && mv /tmp/juicefs /bin/juicefs
   note

   Replace /path/to/kubectl and /path/to/new/juicefs in the script with the actual values, then execute the script.

2. Restart the applications one by one, or kill the existing pods.