Ceph » rbd

From https://github.com/ceph/ceph/pull/55234#issuecomment-1899290071,
"
After a RBD image is mapped to a NBD device using the netlink interface, resizing of the image doesn't update the number of blocks of the NBD device (as seen in /proc/partitions). Only after the device is un-mapped and the image is re-mapped using the netlink interface, the correct block device size of the NBD device is reported.

$ rbd --cluster=site-a -p data create img-use-netlink --size 64M
$ sudo ./bin/rbd --cluster=site-a device map -t nbd -o try-netlink data/img-use-netlink
/dev/nbd4
$ awk -v dev="nbd4" '$4 == dev {print $3}' /proc/partitions
65536
$ rbd --cluster=site-a resize data/img-use-netlink --size 128M
 Resizing image: 100% complete...done.
$ awk -v dev="nbd4" '$4 == dev {print $3}' /proc/partitions
 65536
$  sleep 60; awk -v dev="nbd4" '$4 == dev {print $3}' /proc/partitions
65536
$ sudo ./bin/rbd --cluster=site-a device unmap -t nbd /dev/nbd4
$ sudo ./bin/rbd --cluster=site-a device map -t nbd -o try-netlink data/img-use-netlink
$ # correct value reported after remapping
$ awk -v dev="nbd4" '$4 == dev {print $3}' /proc/partitions
131072

"

In /var/log/messages observed the following after trying to resize,

localhost kernel: block nbd4: must specify backend

Hitting the following error in "drivers/block/nbd.c" kernel code,
https://github.com/ceph/ceph-client/blob/ceph-for-6.8-rc1/drivers/block/nbd.c#L2233

Observed in `netlink_resize()` of tools/rbd_nbd/rbd_nbd.cc that the backend device identifier (cookie) is not specified in the message sent to kernel to resize
https://github.com/ceph/ceph/blob/v19.0.0/src/tools/rbd_nbd/rbd-nbd.cc#L1349