Very strange stripeset problem

From: <marty.cruchten_at_paulwurth.com>
Date: Thu, 30 Sep 1999 17:41:08 +0200

Hello managers

I have a strange problem and I am not able to locate it!
Here the description of the problem:

I have 2 HSZ40 controllers with single disks and stripesets
configured on the controller. The different units are
mirrored through LSM. I used to have a unit D102 on the HSZ40
made up of a stripeset S1 containing DISK120 and DISK410
which are 2 RZ29B (4GB) disks. This unit has become free
because it was replaced by a stripeset of 3 disks. Now I want
to reuse the free disks with 2 additional RZ29B (DISK540
and DISK630) to form a new stripeset S1 containing 4 disks.

I deleted the unit and stripeset S1 on the controller and
recreated the stripeset, initialized the stripeset and recreated
unit D102.

Unit D102 was and is accessible under the name rzc9 on Unix level.
I did the following in UNIX:

> disklabel -z rzc9
> disklabel -rw rzc9 HSZ40
> disklabel -r rzc9

# /dev/rrzc9a:
type: SCSI
disk: HSZ40
label:
flags: dynamic_geometry
bytes/sector: 512
sectors/track: 113
tracks/cylinder: 20
sectors/cylinder: 2260
cylinders: 14828
sectors/unit: 33512112
rpm: 3600
interleave: 1
trackskew: 7
cylinderskew: 26
headswitch: 0 # milliseconds
track-to-track seek: 0 # milliseconds
drivedata: 0

8 partitions:
# size offset fstype [fsize bsize cpg]
  a: 131072 0 unused 0 0 # (Cyl. 0 - 57*)
  b: 262144 131072 unused 0 0 # (Cyl. 57*- 173*)
  c: 33512112 0 unused 0 0 # (Cyl. 0 -
14828*)
  d: 0 0 unused 0 0 # (Cyl. 0 - -1)
  e: 0 0 unused 0 0 # (Cyl. 0 - -1)
  f: 0 0 unused 0 0 # (Cyl. 0 - -1)
  g: 33118896 393216 unused 0 0 # (Cyl. 173*-
14828*)
  h: 0 0 unused 0 0 # (Cyl. 0 - -1)


This disklabel seems OK and shows the total size (4*4=16GB) of the
stripeset.

I tried now to integrate the disk rzc9 into LSM with voldiskadm. voldiskadm
reports the following error :

> Initialize a disk
> Menu: LogicalStorageManager/Disk/AddDisk
>
> Initialization of disk device rzc9 failed.
> Error: voldisk: Device rzc9: define failed:
> Disk write failure

After this command, the label shows as follows:

# /dev/rrzc9a:
type: SCSI
disk: HSZ40
label:
flags: dynamic_geometry
bytes/sector: 512
sectors/track: 113
tracks/cylinder: 20
sectors/cylinder: 2260
cylinders: 14828
sectors/unit: 33512112
rpm: 3600
interleave: 1
trackskew: 7
cylinderskew: 26
headswitch: 0 # milliseconds
track-to-track seek: 0 # milliseconds
drivedata: 0

8 partitions:
# size offset fstype [fsize bsize cpg]
  a: 131072 0 unused 0 0 # (Cyl. 0 - 57*)
  b: 262144 131072 unused 0 0 # (Cyl. 57*- 173*)
  c: 33512112 0 unused 0 0 # (Cyl. 0 -
14828*)
  d: 0 0 unused 0 0 # (Cyl. 0 - -1)
  e: 0 0 unused 0 0 # (Cyl. 0 - -1)
  f: 0 0 unused 0 0 # (Cyl. 0 - -1)
  g: 16362840 393216 unused 0 0 # (Cyl. 173*-
7414*)
  h: 0 0 unused 0 0 # (Cyl. 0 - -1)


LSM seems not to be able to write to the area 8-16GB.

My further tests reveiled that, also advfs (not using LSM) has the same
problem. I initialized the disk with the following partition table:

# size offset fstype [fsize bsize cpg]
  a: 131072 0 unused 0 0 # (Cyl. 0 - 57*)
  b: 262144 131072 unused 0 0 # (Cyl. 57*- 173*)
  c: 33512112 0 unused 0 0 # (Cyl. 0 -
14828*)
  d: 0 0 unused 0 0 # (Cyl. 0 - -1)
  e: 0 0 unused 0 0 # (Cyl. 0 - -1)
  f: 0 0 unused 0 0 # (Cyl. 0 - -1)
  g: 16362840 393216 unused 0 0 # (Cyl. 173*-
7414*)
  h: 16756056 16756056 unused 0 0 # (Cyl. 0 - -1)

Creating an advfs file domain on partition g: or c: caused NO problem. I
could write
data to the stripeset and could see that all 4 disks were used. Creating one
on
partition h: caused the following error message:

mkfdmn /dev/rzc9h testdom
mkfdmn: can't get device size; can't figure out file system partition
mkfdmn: can't create new domain 'testdom'


Then, I tried do map the stripeset S1 on the HSZ40 to another unit: D401. I
didn't use this unit until now. This unit name maps to device rzb12 on UNIX
level.
NOW, with this device name, there were no problems at all for LSM to
initialize the disks and take the whole 16GB for its public region (expect
the little
private region h: of 512 sectors)

So, I first thought it could be an LSM problem, but AdvFS had the same
problem. If it
would be a HSZ40 problem, how could it be explained because I am able to
write to the
first 8GB of the stripeset?

I have no idea at all. If you have any, please help. Don't hesitate to ask
me
more details.


Best regards

Marty Cruchten
Paul Wurth S.A.

marty.cruchten_at_paulwurth.com

   
Received on Thu Sep 30 1999 - 15:43:42 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:39 NZDT