SUMMARY: LSM, voldg import fails

From: Hines, Bruce D (bruce.hines@eds.com)
Date: Tue Feb 25 2003 - 13:12:04 EST


Thanks to Lucien Hercaud.

The issues was that LSM vold automatically configures disk devices that can
be found by inspecting the kernel disk drivers. These auto_configured disk
devices are not stored in persistent configurations, but are regenerated
from kernel tables are every reboot. When the system was rebooted and vold
performed auto_configuration it would detect sliced disks belonging to
another system on the share scsi bus. The disk devices were marked as
online but with a flag called "aliased", which would not allow the disk to
be used. Basically a way for LSM to protect itself from attempting to use
disks improperly. Within TruCluster this would not have been a problem, but
using shared scsi disks without TruCluster this caused a problem when moving
a LSM diskgroup between systems. To fix the problem, the parameter "-x
noautoconfig" was added to the vold daemon when it started.

        Environment:
                5 x GS140 using EMC Symmetrix (scsi)
                Tru64 v5.1A (pk3)
                using scsi hubs to provide shared buses
                NO TruCluster (poor man's cluster)
                LSM dg represent the applications, failover occurs manually
via scripts

        Situation:
                All five systems were upgraded from v4.0x to v5.1A. The
first attempted manual failover was unsuccessfull. LSM said disk volumes
were not usable on the failover host, worked fine on originating host. The
following messages appear on fail to host when booted. This represnt all
the disk volumes in two different dg, one of which was our test failover.
Doing a voldisk list on the fail to host yields following:

                dsk61 sliced - - online
aliased

                same for all the disk volumes in the list below. The first
thing attempted was to have EMC turn on the C-bit to indicate multiple
paths. When we use hwmgr to show wwid for same disk, it has different wwid
on each host. Turning on C-bit did fix that but created all new disks. The
"rejected as clone" still appeared and disk volumes were not usable.

        Solutions:
                We know the real solution is for our customer to purchase
TruCluster (and they know it) but we are attempting to get a poor man's
cluster to work in v5.1A. BTW, this was a working poor man's cluster in
v4.0F.

        starting LSM in boot mode
        lsm:vold: WARNING: Disk dsk61: Disk rejected as clone
        .
        .
        .
        lsm:vold: WARNING: Disk dsk18: Disk rejected as clone

Bruce Hines
Systems Architect
EDS GOSD Software Services
AI TechEng Technology Planning
*+01-405-771-5970 Office
*+01-405-210-8230 Mobile
*aim: BruceHi53
* <mailto:bruce.hines@eds.com>



This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:09 EDT