Hwmgr -scan scsi crashed several systems in my StorageWorks (HSG8 0)

From: Stephane.Moser@gs-vbs.admin.ch
Date: Mon Mar 31 2003 - 02:37:24 EST


Hi everybody

I had last Friday a very strange problem with my StorageWorks.

My HSG80 (Software V86F-4) configuration look like that:

stripeset DVGRPSM4
MIRR_16 (DISK10300,DISK20300)
MIRR_15 (DISK30300,DISK40300)
MIRR_14 (DISK50300, DISK60300)

Host_1 18.2GB (Partition on stripeset DVGRPSM4)
Host_2 18.2GB (Partition on stripeset DVGRPSM4)
Host_3 18.2GB (Partition on stripeset DVGRPSM4)
Host_4 18.2GB (Partition on stripeset DVGRPSM4)
Host_5 18.2GB (Partition on stripeset DVGRPSM4)
Host_6 18.2GB (Partition on stripeset DVGRPSM4)

The root_domain, usr_domain, and var_domain are on Storageworks and the
data_domain are on Hitachi disk subsystem.

I have naturally other stripsets but I didn't notice any problem with them.

I modified the /etc/ddr.dbase (see below) file on the Host_2 (Tru64 5.1A) to
have the best configuration for my Hitachi 9980V disk subsystem. After I
changed this file I made a ddr_config -c to activate this change.

I made an hwmgr -scan scsi on the Host_2 to see my new devices and this
command crashed the Host_1 (Tru64 5.1), Host_3 (Tru64 5.1) and Host_4 (Tru64
5.1). It was impossible to restart Host_1, Host_3 and Host_4 (wwidmgr: not
connected devices) I had to boot first with the CD-ROM and only after it was
possible to boot normally. After that I restarted the Host_2 and during the
boot process 4 Systems crashed. I remove my modifications in the
/etc/ddr.dbase file and everything work fine.

I have 3 questions:

1) Are my modifications in the /etc/ddr.dbase file so "dangerous"?
2) It's really possible inside a StorageWorks from one system to crash
other systems?
3) Where is exactly the problem Host_2 or HSG80 (StorageWorks)?

Any hints are really welcome.

Best regards,

Stephane

***************************************************************
(White paper from Hitachi)

/etc/ddr.dbase

SCSIDEVICE
    #
    #
    Type = disk
    NAME = "DEC" "HSG80"
    #
    PARAMETERS:
        TypeSubClass = hard_disk, raid
        BadBlockRecovery = disabled
        TagQueueDepth = 32
        DisperseQueue = false
        RequestSenseLength = 96
    ATTRIBUTE:
        AttributeName = "VPDinfo"
        Length = 16
        Ubyte [0] = 0x00
        Ubyte [1] = 0x00
        Ubyte [2] = 0x00
        Ubyte [3] = 0x00
        Ubyte [4] = 0x00
        Ubyte [5] = 0x0f
        Ubyte [6] = 0x00
        Ubyte [7] = 0x00
        Ubyte [8] = 0x00
        Ubyte [9] = 0x00
        Ubyte [10] = 0x00
        Ubyte [11] = 0x00
        Ubyte [12] = 0x00
        Ubyte [13] = 0x00
        Ubyte [14] = 0x00
        Ubyte [15] = 0x00

***************************************************************

Stéphane Moser
DIK-VBS Abteilung Rechenzentrum
Sektion Datenmanagement
Papiermühlestrasse 20
3003 Bern
Tel: +41 31 32 34294
mailto:stephane.moser@gs-vbs.admin.ch



This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:14 EDT