After Reboot sg path does not exist

[piscu]

Hi, I have one Robot defined and two Drives in my NB configuration. We have upgraded to nb7.1 and after the upgrade all was up and running and I could see my Robot and the two Drives: /usr/openv/volmgr/bin/sgscan /dev/sg/c0tw50014380032c1d5al0: Tape (/dev/rmt/0): "HP Ultrium 4-SCSI" /dev/sg/c0tw50014380032c1d5al1: Changer: "HP MSL G3 Series" /dev/sg/c0tw50014380032c1d60l0: Tape (/dev/rmt/1): "HP Ultrium 4-SCSI" But after a reboot I get all the time the following error message in the console: tldcd[1097]: TLD(1) [1097] robotic path /dev/sg/c0tw50014380032c1d5al1 does not exist tldd[1093]: TLD(1) unavailable: initialization failed: Unable to open robotic path It seems that after a reboot the raw file is missing where the link /dev/sg/c0tw50014380032c1d5al1 is pointing to: ls -la /dev/sg/c0tw50014380032c1d5al1 lrwxrwxrwx 1 root root 85 Apr 8 11:59 /dev/sg/c0tw50014380032c1d5al1 -> ../../devices/pci@0,0/pci8086,340c@5/pci10df,fc42@0/fp@0,0/sg@w50014380032c1d5a,1:raw Workaround: I can only solve the problem when I do a "ls -la /dev/sg/c0tw50014380032c1d5al1". At that moment the raw file is generated. Does anybody know why this happens after a reboot? I tried several times to recreate the sg.config with "sg.install", but it did not help... As additional Information: - SW: Solaris 10 9/10 (x86) - HW Server: Sun Fire x4270 M2 - HW Robot: Sun SL48 - NB Version: NetBackup 7.1 Thanks and br

[Marianne]

Please verify that persistent binding is in place to ensure that device addresses do not change when system is rebooted.

What hba and driver is used in your master ? (check /var/adm/messages shortly after reboot)

What does 'sgscan' and 'scan' report after the reboot?

[piscu]

After each reboot I get the same device address... It seems that the raw file is missing where the device address is pointing to. Here are the messages after the last reboot: ... Apr 15 10:14:18 hsrvbkpt1 scsi: [ID 799468 kern.info] st3 at fp0: name w50014380032c1d5a,0, bus address 2 Apr 15 10:14:18 hsrvbkpt1 genunix: [ID 936769 kern.info] st3 is /pci@0,0/pci8086,340c@5/pci10df,fc42@0/fp@0,0/tape@w50014380032c1d5a,0 Apr 15 10:14:18 hsrvbkpt1 scsi: [ID 365881 kern.info] /pci@0,0/pci8086,340c@5/pci10df,fc42@0/fp@0,0/tape@w50014380032c1d5a,0 (st3): Apr 15 10:14:18 hsrvbkpt1 Apr 15 10:14:19 hsrvbkpt1 scsi: [ID 799468 kern.info] st2 at fp2: name w50014380032c1d60,0, bus address 2 Apr 15 10:14:19 hsrvbkpt1 genunix: [ID 936769 kern.info] st2 is /pci@0,0/pci8086,340c@5/pci10df,fc42@0,1/fp@0,0/tape@w50014380032c1d60,0 Apr 15 10:14:19 hsrvbkpt1 scsi: [ID 365881 kern.info] /pci@0,0/pci8086,340c@5/pci10df,fc42@0,1/fp@0,0/tape@w50014380032c1d60,0 (st2): Apr 15 10:14:19 hsrvbkpt1 Apr 15 10:14:26 hsrvbkpt1 scsi: [ID 799468 kern.info] sgen0 at scsi_vhci0: name g50014380032c1d58, bus address g50014380032c1d58 Apr 15 10:14:26 hsrvbkpt1 genunix: [ID 936769 kern.info] sgen0 is /scsi_vhci/medium-changer@g50014380032c1d58 Apr 15 10:14:26 hsrvbkpt1 tldcd[1097]: [ID 295976 daemon.error] TLD(1) [1097] robotic path /dev/sg/c0tw50014380032c1d5al1 does not exist Apr 15 10:14:26 hsrvbkpt1 last message repeated 1 time Apr 15 10:14:26 hsrvbkpt1 tldd[1093]: [ID 915677 daemon.error] TLD(1) unavailable: initialization failed: Unable to open robotic path Apr 15 10:14:27 hsrvbkpt1 avrd[1094]: [ID 920050 daemon.notice] st.conf configuration for r1_drive0 (device 0), name [HP Ultrium LTO 4], vid [HP Ultrium 4*], type 0x3b, block size 0, options 0x18619 (see st(7D) man page) Apr 15 10:14:27 hsrvbkpt1 sg: [ID 266374 kern.notice] Symantec SCSA Generic Revision: 3.7 Apr 15 10:14:27 hsrvbkpt1 avrd[1094]: [ID 262211 daemon.notice] st.conf configuration for r1_drive1 (device 1), name [HP Ultrium LTO 4], vid [HP Ultrium 4*], type 0x3b, block size 0, options 0x18619 (see st(7D) man page) Apr 15 10:16:28 hsrvbkpt1 tldcd[1097]: [ID 295976 daemon.error] TLD(1) [1097] robotic path /dev/sg/c0tw50014380032c1d5al1 does not exist Apr 15 10:16:28 hsrvbkpt1 last message repeated 1 time Apr 15 10:16:28 hsrvbkpt1 tldd[1093]: [ID 915677 daemon.error] TLD(1) unavailable: initialization failed: Unable to open robotic path Apr 15 10:18:30 hsrvbkpt1 tldd[1093]: [ID 947741 daemon.notice] TLD(1) going to UP state Apr 15 10:38:50 hsrvbkpt1 pseudo: [ID 129642 kern.info] pseudo-device: fcsm0 Apr 15 10:38:50 hsrvbkpt1 genunix: [ID 936769 kern.info] fcsm0 is /pseudo/fcsm@0 Apr 15 10:38:51 hsrvbkpt1 scsi: [ID 193665 kern.info] sd2 at mr_sas0: target 7 lun 0 Apr 15 10:38:51 hsrvbkpt1 genunix: [ID 936769 kern.info] sd2 is /pci@0,0/pci8086,340a@3/pci1000,9263@0/sd@7,0 ... And 'sgscan' or 'scan' are just reporting the two drives. The Robot is missing.

[Marianne]

Have a look at the device address that the OS sees and the device name configured in NBU:

OS medium-changer: 50014380032c1d58
expected sg device:   50014380032c1d5a

Try to rebuild sg drivers:

cd /usr/openv/volmgr/bin/driver
# /usr/openv/volmgr/bin/sg.build all -mt <15> -ml <1>

Install the new sg driver configuration:

# /usr/bin/rm -f /kernel/drv/sg.conf
# /usr/openv/volmgr/bin/driver/sg.install

Check/verify config:

# /usr/openv/volmgr/bin/sgscan

[piscu]

I tried it but it put again "50014380032c1d5a" and 'sgscan' didn't show the Robot after the sg.install. I made the following: # cd /usr/openv/volmgr/bin/driver # /usr/openv/volmgr/bin/sg.build all -mt 15 -ml 1 # more sg.conf # Configuration file for SCSA Generic. # name="sg" class="scsi" target=0 lun=0; name="sg" class="scsi" target=0 lun=1; name="sg" class="scsi" target=1 lun=0; name="sg" class="scsi" target=1 lun=1; name="sg" class="scsi" target=2 lun=0; name="sg" class="scsi" target=2 lun=1; name="sg" class="scsi" target=3 lun=0; name="sg" class="scsi" target=3 lun=1; name="sg" class="scsi" target=4 lun=0; name="sg" class="scsi" target=4 lun=1; name="sg" class="scsi" target=5 lun=0; name="sg" class="scsi" target=5 lun=1; name="sg" class="scsi" target=6 lun=0; name="sg" class="scsi" target=6 lun=1; name="sg" class="scsi" target=8 lun=0; name="sg" class="scsi" target=8 lun=1; name="sg" class="scsi" target=9 lun=0; name="sg" class="scsi" target=9 lun=1; name="sg" class="scsi" target=10 lun=0; name="sg" class="scsi" target=10 lun=1; name="sg" class="scsi" target=11 lun=0; name="sg" class="scsi" target=11 lun=1; name="sg" class="scsi" target=12 lun=0; name="sg" class="scsi" target=12 lun=1; name="sg" class="scsi" target=13 lun=0; name="sg" class="scsi" target=13 lun=1; name="sg" class="scsi" target=14 lun=0; name="sg" class="scsi" target=14 lun=1; name="sg" class="scsi" target=15 lun=0; name="sg" class="scsi" target=15 lun=1; name="sg" parent="fp" target=0 lun=0 fc-port-wwn="50014380032c1d5a"; name="sg" parent="fp" target=0 lun=1 fc-port-wwn="50014380032c1d5a"; name="sg" parent="fp" target=0 lun=0 fc-port-wwn="50014380032c1d60"; name="sg" parent="fp" target=0 lun=1 fc-port-wwn="50014380032c1d60"; # /usr/bin/rm -f /kernel/drv/sg.conf # /usr/openv/volmgr/bin/driver/sg.install Copied files to /kernel/drv/amd64. Doing add_drv of the sg driver Removing old /dev/sg entries. Editing /etc/devlink.tab... Copying original /etc/devlink.tab to /etc/devlink.tab.04-19-11-07:23:49. Added entry in /etc/devlink.tab file. Made links in /dev/sg # /usr/openv/volmgr/bin/sgscan /dev/sg/c0tw50014380032c1d5al0: Tape (/dev/rmt/0): "HP Ultrium 4-SCSI" /dev/sg/c0tw50014380032c1d60l0: Tape (/dev/rmt/1): "HP Ultrium 4-SCSI"

[Marianne]

I am not too familiar with Solaris x86.

I'm wondering if the the Solaris sgen driver is preventing the sg driver from using the medium changer.

There is also a Solaris patch that 'breaks' the sg driver - 127128-11 (x86), although you are not seeing the errors described in http://www.symantec.com/docs/TECH56850 .

Please try to disable/unload the Solaris sgen driver and retry sg.build and sg.install.

[piscu · Accepted solution]

Thanks for the answers... I think I found the answer in the NetBackup Device Config Guide: There it's written that NetBackup requires that Solaris Multiplexed I/O (MPxIO) be disabled. MPxIO is enabled by default on Solaris 10 x86 systems. Therefore I disabled the multiple paths to the same tape drive: vi /kernel/drv/fp.conf ... # mpxio-disable="no" mpxio-disable="yes" Maybe it helps others too...