Networker ejects only drives that don't have lus drivers

[Mario2]

Hi,


I have a case with Storagetek L700e that has 6 SDLT tape drives connected over 3 SCSI to fiber routers (SN3250)to a switch.

The Legato lus_add_fp_devs script adds only 3 lus instances as it only sees the 3 routers and not 6 drives. So the inquire command sees only 3 tape devices.

The jbconfig command sees and configures all 6 SDLT drives, although the number of lus instances still stays the same.

The problem is that Networker is working properly with the tape drives that are not listed with inquire (the ones that don’t have lus instance),

but after finishing the backup on any of other 3 drives (the ones the inquire sees) the tape will not be ejected.

The eject operation fails, but if I issue the mt –f /dev/rmt/name_of_the_drive_loaded offline during the eject request, the tape is rewound and returned successfully to its origin slot. The backup finishes OK then.

Is there a reason that Networker would work properly with the drive (load, position, write,read) but not eject the tape.

And that is only happening on the B side of router ports. (swapping the drives did not work)

I tried manually editing the lus.conf file to enable inquire to see all 6 drives, but the results are more or less the same. Is it essential for Networker to have all 6 lus files? That is important to me so I can continue troubleshooting in the right direction.

This hole situation occurred with the upgrade of Networker from 6 to 7.1.1.

Thank you very much for any info!

Mario

 

[nohandleisworking]

Hi,

(1) make sure you have the newest lus.conf file from legato:

http://www.legato.com/resources/bulletins/390.html

I have pasted most recent one i could find below. Also if you have access to legato web support there is an excellent article I remember reading not too long ago on understanding and configuring lus.conf.

Also, look in Networker Unix Admin guide for 7.1 starting on page 246 -248 on lus.conf.
also, check what the command lusbinfo and "lusbinfo -v" are saying you have.
In my lus file it is basically the same as the original except there is one line added at the very end under
name="lus" class="scsi" target=0 lun=0;

I basically have just one line with the target and lun and correct hba for the "autochanger" only of the library on the SAN. I did not need to list a different "name" line for all of my many drives. lus (legato user scsi) controls the robotic arm of library so the not being able to eject is probably related since legato needs to be configured right so it can talk to the library and tell it what it needs to do.






/usr/kernel/drv > cat lus.conf
#
# $Id: lus.conf.src,v 1.12.52.1.8.3.12.1 2004/03/29 21:25:55 jquintei Exp $ Copyright (c) 2004, Legato Systems, Inc.
#
# Copyright (c) 2004 LEGATO Software, a division of EMC.
#
# All rights reserved. This is an UNPUBLISHED work, and
# comprises proprietary and confidential information of LEGATO.
# Unauthorized use, disclosure, and distribution are strictly
# prohibited. Use, duplication, or disclosure of the software
# and documentation by the U.S. Government is subject to
# restrictions set forth in a license agreement between the
# Government and LEGATO or other written agreement specifying
# the Government's rights to use the software and any applicable
# FAR provisions, such as FAR 52.227-19.
#

# Configuration for number of targets on scsi busses: new with v 6.12 of lus.
# There are 3 methods lus uses to determine how many target IDs to allow for a
# given SCSI bus.
#
# First, the list below in the property called known-scsi-adapters is searched
# for a match with the SCSI adapter's device name. If an exact match is found,
# the number after the : is used as the number of targets. Each entry is
# separated by at least one space. The number after the - is the maximum LUN
# that will be searched. If there is a character 'C' as the last character
# before the next space, then lus will do what is called HBA transport cloning.
# If you don't know a driver requires this. do not set it as we suspect that
# some normal SCSI HBA drivers can cause system panics if we do this.
# The usual driver that needs cloning is a Fibrechannel driver that does some
# sort of clever mapping between Fibrechannel LUNs and SCSI IDs
#
# If there is no match in known-scsi-adapters, another property called
# bus<n>-targets is searched for. If a match is found for the current bus
# number, the value after the = is used for the number of targets on that
# bus. (e.g., for bus 1, if bus1-targets=16; is found, then 16 will be used
# for bus 1) To use this method, simply insert this type of string in this
# file below where indicated for global properties. Usual values for
# bus<n>targets will be 8 for narrow SCSI, 16 for wide SCSI or 126 for
# FiberChannel adapters with SCSI-like drivers
#
# If none of the above set a value, lus looks for a property called "wide-xfer".
# If this is found on a given adapter, then 16 targets is assumed, otherwise 8
# is used.
#
# There will be one line displayed on the system console when lus loads for each
# bus found, stating the number of targets being used and the method used to
# determine the value.
#
# Warning: Setting these values too high can have dire consequences for your
# data. Experimentation is at the experimenter's own risk.
#
# Global properties recognized include:
# known-scsi-adapters="adapter_name:target_count-maxlun[C] adapter_name:target_count-maxlun[C] ...";
# bus<n>-targets=X;
# bus<n>-maxlun=X;
#
#

# Insert global properties here, (one per line, ending with a semicolon)
# before the 'name="lus"' line.

# esp = Sun SE scsi I/F
# isp = SUN Differential or fast/wide SCSI
# fas = Sun fast wide SCSI
# glm = Sun SCSI in Ultras
# pci1000,f = alternate driver for glm PCI adapters
# ifp = (don't recall - is some FC adapter)
# pci1077,2100 = Qlogic 2100 PCI Fibrechannel
# pci1077,1 = Qlogic 2100 - different rev driver
# pci1077,2200 = Qlogic 2200, rev 1.09 driver
# pci9004,8478 = Adaptec 2944 UW/OF differential scsi
# pci10df,f700 = Emulex LP7000 Fibrechannel (needs clone)
# pci10df,f800 = Emulex LP8000 Fibrechannel (needs clone)
# pci10df,f900 = Emulex LP9000 Fibrechannel (needs clone)
# fca = JNI 32bit Fibrechannel (needs clone)
# fcaw = JNI 64bit Fibrechannel (needs clone)
# pci1242,4643 = JNI PCI 64-bit Fibrechannel (needs clone)
# JNI,FCE = JNI PCI 64-bit Fibrechannel (needs clone)
# fce = Driver 5.3.x for above
# JNI,FCR = JNI PCI 64-bit Fibrechannel (needs clone)
# fcr = Driver 5.3.x for above
# fp = Sun's new "leadville" FC driver stack on Solaris 8 only
# pci1077,2 = Qlogic 2200, rev 3.07 driver
# pci1077,9 = Qlogic 2300 rev 4.03 driver
# pci10cf,105c = Fujitsu PFCA adapter
# pci10cf,1178 = Fujitsu e-PFCA adapter (2G)
# fjulsa = Fujitsu LVD320 adapter

known-scsi-adapters="esp:8-7 SUNW,esp:8-7 isp:16-7 SUNW,isp:16-7 QLGC,isp:16-7 fas:16-7 SUNW,fas:16-7 glm:16-7 SUNW,glm:16-7 pci1000,f:16-7 ifp:126-126 SUNW,ifp:126-126 PTIX,isptwo:126-126 pci1077,2100:126-126 pci1077,1:126-126 pci9004,8478:16-7 pci10df,f700:126-32C pci10df,f800:126-32C pci10df,f900:126-32C fca:126-32C fcaw:126-32C pci1242,4643:126-32C pci1077,2200:126-32 JNI,FCE:126-32C fce:126-32C JNI,FCR:126-32C fcr:126-32C fp:126-32 pci1077,2:126-32 pci1077,9:126-32 pci10cf,105c:126-32C pci10cf,1178:126-32C fjulsa:16-7 ";

name="lus" class="scsi" target=0 lun=0;

 

[Mario2]

Thanks for your answer,

Unfortunately the problem was in something quite different.
I had to disable the usage of CDI interface in NetWorker by creating a 0 Byte file
touch /nsr/debug/cdidisable
Restarted the NetWorker daemons, checked for all tape device to show up option "CDI: not used", it didn’t so I changed it manually.
It seems that the GUI does not reflect the /nsr/debug/cdidisable usage but it works even without the GUI set – just using the /nsr/debug/cdidisable file.

The CDI is a new module that came with Networker 7.1.x, apparently it has some major bugs in it because as soon as I disabled that thing everything started to work properly.