Skip to main content

FAQ

My question is not answered in the documentations​

Try searching for your problems in the top right corner, using different keywords. If nothing comes up:

How to seamlessly remount JuiceFS file system?​

If you can accept downtime, simply delete the Mount Pod and JuiceFS is remounted when the Mount Pod is re-created (note that if automatic mount point recovery isn't enabled, you'll need to restart or re-create application Pods to bring mount point back into service). But in Kubernetes, we often wish a seamless remount. You can achieve a seamless remount by the following process:

  • When upgrading or downgrading CSI Driver, if the Mount Pod image is changed along the way, CSI Driver will create a new Mount Pod when you perform a rolling upgrade on application Pods.
  • Modify mount options at PV level, and perform a rolling upgrade on application Pods. Note that for dynamic provisioning, although you can modify mount options in StorageClass, but the changes made will not be reflected on existing PVs, a rolling upgrade thereafter will not trigger Mount Pod re-creation.
  • Modify volume credentials, and perform a rolling upgrade on application Pods.
  • If no configuration has been modified, but a seamless remount is still in need, you can make some trivial, ineffective changes to mount options (e.g. increase cache-size by 1), and then perform a rolling upgrade on application Pods.

To learn about the CSI Driver implementation and find out when new Mount Pods will be created to achieve seamless remount, see the GenHashOfSetting() function in pkg/juicefs/mount/pod_mount.go.

exec format error​

The most common cause is CPU architecture misalignment, like running ARM64 images on a x86 system. If you must transfer images using a personal computer of different architecture, remember to specify the platform:

# Use the architecture of the actual environment
docker pull --platform=linux/amd64 juicedata/mount:ee-5.0.17-0c63dc5

Apart from architecture misalignment, we've also seen this type of errors caused by bad container runtime, in which case exec format error can occur even if images are of the same architecture. When this happens, you might need to purge and reinstall container runtime, take containerd as an example:

systemctl stop kubelet
rm -rf /var/lib/containerd
# reinstall containerd
Follow us on X
Let's chat on Slack