"sdx device busy" Displayed When Creating Bcache on Ceph

Symptom

In a Ceph environment configured with bcache, reinstalling the operating environment may occasionally cause an error during bcache configuration, stating "make-bcache --wipe-bcache -B /dev/sdx -C /dev/nvmexx". The details are shown in the screenshot below:

Key Process and Cause Analysis

Residual bcache partition information on the drive is loaded by the system and can be checked using the lsblk command.

Conclusion and Solution

Perform the following steps to recover the drive:

Perform detach and stop operations in the bcache folder under the sdb directory.
```
echo 1 >/sys/block/sdb/bcache/detach
echo 1 >/sys/block/sdb/bcache/stop
```

Perform a dd write operation on the sdb drive.

dd if=/dev/zero of=/dev/sdb count=2048 bs=1M

Unregister the residual bcache under the /sys/fs/bcache/ directory.
```
cd  /sys/fs/bcache/31040c7b-55f2-4170-bf77-d2e8b958e31a/
echo 1 >  unregister
```
In the corresponding bcache directory, perform an unregister operation.
Reboot the server.

Script reference:

Perform detach, stop, and dd write operations on drives sda through sdh.

for i in {0..7}
do
        temp=`expr $i % 8 + 97`
        diskID=$(printf \\x`printf %x ${temp}`)
        echo 1 >/sys/block/sd${diskID}/bcache/detach
        echo 1 >/sys/block/sd${diskID}/bcache/stop
        dd if=/dev/zero of=/dev/sd${diskID} count=2048 bs=1M
done

Perform the unregister operation on the /sys/fs/bcache/ directory across nodes ceph1 through ceph4.

#!/bin/bash
for n in {1..4}
do
    echo "ceph$n"
    f=$(ssh ceph$n "ls /sys/fs/bcache/")
    for i in $f
    do
        ssh ceph$n "echo 1 > /sys/fs/bcache/$i/unregister"
    done
done