multipath devices not reflected as partitioned

SUSE Product Topics SLES Configure-Administer
1

Hi,

here’s my investigation


NODE01:~ # multipath -ll ocr-vote-mirror-1
ocr-vote-mirror-1 (3600507630080828c4000000000000016) dm-15 IBM     ,2145
size=500M features='1 queue_if_no_path' hwhandler='0' wp=rw
`-+- policy='round-robin 0' prio=30 status=active
  |- 7:0:0:20 sdv  65:80  active ready running
  `- 8:0:0:20 sdas 66:192 active ready running
  
NODE01:~ # /lib/udev/scsi_id --whitelisted --device=/dev/sdv
3600507630080828c4000000000000016
NODE01:~ # /lib/udev/scsi_id --whitelisted --device=/dev/sdas
3600507630080828c4000000000000016

NODE01:~ # ls -l /dev/mapper/ocr-vote-mirror-1*
lrwxrwxrwx 1 root root 8 Apr 18 15:14 /dev/mapper/ocr-vote-mirror-1 -> ../dm-15
lrwxrwxrwx 1 root root 8 Apr 18 09:08 /dev/mapper/ocr-vote-mirror-1_part1 -> ../dm-46
NODE01:~ # ls -l /dev/sdv*
brw-rw---- 1 root disk 65, 80 Apr 17 14:06 /dev/sdv
NODE01:~ # ls -l /dev/sdas*
brw-rw---- 1 root disk 66, 192 Apr 17 14:06 /dev/sdas

At node01, both /dev/sdv and /dev/sdas are not reflected as partitioned.

  
NODE02:~ # multipath -ll ocr-vote-mirror-1
ocr-vote-mirror-1 (3600507630080828c4000000000000016) dm-15 IBM     ,2145
size=500M features='1 queue_if_no_path' hwhandler='0' wp=rw
`-+- policy='round-robin 0' prio=30 status=active
  |- 7:0:0:20 sdv  65:80  active ready running
  `- 8:0:0:20 sdas 66:192 active ready running
  
NODE02:~ # /lib/udev/scsi_id --whitelisted --device=/dev/sdv
3600507630080828c4000000000000016
NODE02:~ # /lib/udev/scsi_id --whitelisted --device=/dev/sdas
3600507630080828c4000000000000016

NODE02:~ # ls -l /dev/mapper/ocr-vote-mirror-1*
lrwxrwxrwx 1 root root 8 Apr 18 15:21 /dev/mapper/ocr-vote-mirror-1 -> ../dm-15
lrwxrwxrwx 1 root root 8 Apr 17 14:51 /dev/mapper/ocr-vote-mirror-1_part1 -> ../dm-32
NODE02:~ # ls -l /dev/sdv*
brw-rw---- 1 root disk 65, 80 Apr 17 14:08 /dev/sdv
NODE02:~ # ls -l /dev/sdas*
brw-rw---- 1 root disk 66, 192 Apr 17 14:08 /dev/sdas

same things happen at node02, both /dev/sdv and /dev/sdas are not reflected as partitioned.

I partitioned by


fdisk /dev/mapper/wwid

from https://www.suse.com/documentation/sled11/stor_admin/data/mpiosysconf.html#mpiosysconfpart

Am I doing the correct thing or wrong things that’s why things don’t turn up as expected?

many thanks in advance.

2

Hi slesmaster,

have you re-attached the devices after partitioning? I wouldn’t expect the system (udevd) to detect the changed partitioning by itself, and fdisk will have notified the kernel only about changing /dev/mapper/wwid…

As a matter of fact, you cannot judge if the servers see the device as partitioned, unless you use i.e. “fdisk -l” to check. What you were looking at is the set of symlinks created by udev, once it detects partitioned devices.

Regards,
Jens

3

I’ve googled reattach multipath devices but don’t seem to know what you mean. Maybe you can enlighten me on this. I’m not a sysadmin, but a DBA with very limited knowledge of Linux OS. I will try to pick up more Linux skills. I may have to tell him what’s wrong and what should be done.

When I used fdisk -l output, the devices is still not reflected as partitioned on either node.


NODE02:~ # fdisk -l /dev/sdv* > /oracle_software/fdisk_output/node02_fdisk_sdv_20140428_0755.log
Disk /dev/sdv doesn't contain a valid partition table
NODE02:~ # fdisk -l /dev/sdas* > /oracle_software/fdisk_output/node02_fdisk_sdas_20140428_0755.log
Disk /dev/sdas doesn't contain a valid partition table

NODE02:~ # cat /oracle_software/fdisk_output/node02_fdisk_sdas_20140428_0755.log

Disk /dev/sdas: 524 MB, 524288000 bytes
17 heads, 59 sectors/track, 1020 cylinders, total 1024000 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000

NODE02:~ # cat /oracle_software/fdisk_output/node02_fdisk_sdv_20140428_0755.log

Disk /dev/sdv: 524 MB, 524288000 bytes
17 heads, 59 sectors/track, 1020 cylinders, total 1024000 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000

NODE02:~ #

same thing over at NODE01.

Is that something wrong with the configuration?

again thanks for your time.

4

Hi slesmaster,

…so your nick is a bit misleading :wink:

I was talking about any situation that will make the device “go away” from SLES’ point of view - the most drastic measure being a reboot. Background to my suggestion is the dynamic nature of the entries you were looking at: These are auto-generated by a subsystem called “udev”, once udev detects a change in devices. Modifying the partition table is not detected automatically.

But:

Thus reattachment will do no good anyhow - if even fdisk -l doesn’t report an appropriate partition table, udev would not be able to create the partition symlinks anyhow.

Allow me to ask if you are aware that fdisk needs you to invoke the “w” subcommand to actually write changes to disk, after having created partitions?

Could you re-run “fdisk /dev/sdv” and create the partitions, save your changes and then, after being on the shell command line again, immediately run “fdisk -l /dev/sdv”? If then the partitions still don’t show, please post a cut&paste from that part so we can look for any non-obvious mistakes…

I won’t be able to respond as quick as usual, at least for a week or two, so if anyone wants to give a helping hand here, too, feel welcome :slight_smile:

Regards,
Jens