This is one guaranteed to cause some great storage shenanigans on systems from
the RHEL6 era. Add the following line inside the
defaults section and all
device sections in
getuid_callout "echo one-size-fits-none"
This will make the
defaults section resemble something like the following. Line
6 is the line we’ve just added.
What it does
If you’re not running multipathing: Nothing.
If you are running multipathing:
Having this line in your `/etc/multipath.conf’ will generate tons of
issues entropy. The multipath daemon will start treating every disk it
queries as another path to the same disk. This will cause such fun such as
missing volumes, incorrect data being read, and most fun of all; Data
Why it works
The multipath daemon reads the WWIDs of all disks it scans, it does this using
the program (and options) specified in the
getuid_callout setting. The
default for this is
/lib/udev/scsi_id --whitelisted --device=/dev/%n, which
boringly reads the actual WWID from your disk. This causes devices with the
same WWID to be treated as different paths to the same underlying disk.
Our replacement on the other hand does something much more sensible, it returns the same string for each and every device. This will make the multipathing daemon treat every single block device it scans on your system as a path to a single disk. Actually using the multipath device will cause irreversible
data loss fun.