I’ve noticed that
multipath devices don’t report manufacturer or model into lsblk, but they do
report model and vendor when running the multipath utility as “multipath –ll”.
See following example (Yes I realize my disks have failed horribly… I was
testing something else at the time.). I have highlighted important areas.
> multipath -ll
mpaths (26664303464626366) dm-3 FUSIONIO,ION LUN
size=93G features='3 queue_if_no_path pg_init_retries 50' hwhandler='1 alua' wp=rw
|-+- policy='queue-length 0' prio=0 status=enabled
| |- 8:0:3:2 sdbg 67:160 failed faulty running
| |- 7:0:3:2 sdas 66:192 failed faulty running
| `- 7:0:2:2 sdac 65:192 failed faulty running
|-+- policy='queue-length 0' prio=0 status=enabled
| |- 7:0:1:2 sds 65:32 failed faulty running
| `- 8:0:0:2 sdh 8:112 failed faulty running
|-+- policy='queue-length 0' prio=0 status=enabled
| |- 7:0:0:2 sdc 8:32 failed faulty running
| `- 8:0:1:2 sdab 65:176 failed faulty running
`-+- policy='queue-length 0' prio=0 status=enabled
`- 8:0:2:2 sdar 66:176 failed faulty running
> lsblk -s -o kname,type,model,vendor /dev/dm-3
KNAME TYPE MODEL VENDOR
dm-3 mpath .
sdc disk ION LUN FUSIONIO
sdh disk ION LUN FUSIONIO
sds disk ION LUN FUSIONIO
sdab disk ION LUN FUSIONIO
sdac disk ION LUN FUSIONIO
sdbg disk ION LUN FUSIONIO
sdar disk ION LUN FUSIONIO
sdas disk ION LUN FUSIONIO
I am assuming that this is because each multipath device is missing a sysfs endpoint for lsblk to read which contains the pertinent model and device information. It would appear that the underlying slave block devices do however have the information.
> cat /sys/class/block/dm-3/slaves/*/device/model
ION LUN
ION LUN
ION LUN
ION LUN
ION LUN
ION LUN
ION LUN
I’ve been trying to find the best place in the code to add the feature myself, but with limited experience, time and other obligations my time for adding additional sysfs endpionts to make lsblk work correctly has been quickly spent up. Maybe I’m going about this the wrong way, but with time short, I wasn’t sure what the best course of action was for subbmitting a feature request or line of questioning of how to add the additional sysfs endpionts without breaking linux as a whole by myself as a novice kernel developer.
Is this something I should even be worrying about, IE. Working as designed? If so, what is the policy on listing the model and volume of a multipath disk?
Should
this issue end up as a fix in lsblk, or the multipath device endpoint in sysfs?
Who should I be senging email to about making this a feature request?
If I truly have to make the fix myself, any advice on where I should be
starting in the multipath code or lsblk code respectivly?
Also, apologies in advance if I am bugging the wrong list about this.
Russell Weber
Software Support and Quality engineer
One Stop Systems
-- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel