Open innobead opened 1 year ago
The underlying issue for many users of Longhorn is that some storage drivers using iSCSI enable multipathd
democratic-csi
enable multipathd? Yes: https://github.com/democratic-csi/democratic-csi?tab=readme-ov-file#iscs
Also, you can see a related issue from AWS EBS CSI: https://github.com/kubernetes-sigs/aws-ebs-csi-driver/issues/1417#issuecomment-1275246361
Seemingly the only thing that needs to be added to the blacklist based on that comment is:
defaults {
user_friendly_names yes
}
blacklist {
device {
vendor "IET"
product "VIRTUAL-DISK"
}
}
because
$ ssh administrator@mynode udevadm info --query=all --name=/dev/sdd | grep -E "ID_MODEL=|ID_VENDOR="
E: ID_VENDOR=IET
E: ID_MODEL=VIRTUAL-DISK
this is the vendor and product name of the longhorn volume
Need to consider the node condition for multipathd
, https://github.com/longhorn/longhorn-manager/pull/2763#discussion_r1671540945
Is your improvement request related to a feature? Please describe (👍 if you like this request)
When running a multipath-enabled environment, users need to exclude Longhorn volumes manually, or else any volume will not be able to attach, because the dev will be managed by multipath. We need to make it automatically handled to provide a better user experience instead of struggling with unnecessary troubleshooting.
Describe the solution you'd like
Describe alternatives you've considered
Manually do https://longhorn.io/kb/troubleshooting-volume-with-multipath/, but before that, users would encounter the attach problem and take time to figure out the cause.
Additional context
cc @longhorn/dev