Tape devices appear as offline after scalar i500 tape library reboot

[Tom Wolf_3]

Hello all.

We have HP-UX 11.31 ia64 hosts that have Scalar i500 tape library drives presented to them.

We noticed that after rebooting the tape library, these tape devices appear as "CLAIMED" but "offline" in ioscan output as shown here.

# ioscan -m lun | grep off
tape      4  64000/0xfa00/0x18   estape  CLAIMED     DEVICE       offline  HP      Ultrium 6-SCSI
tape      5  64000/0xfa00/0x19   estape  CLAIMED     DEVICE       offline  HP      Ultrium 6-SCSI
tape      6  64000/0xfa00/0x1a   estape  CLAIMED     DEVICE       offline  HP      Ultrium 6-SCSI
tape      7  64000/0xfa00/0x1b   estape  CLAIMED     DEVICE       offline  HP      Ultrium 5-SCSI

# ioscan -m lun | grep tape
tape      4  64000/0xfa00/0x18   estape  CLAIMED     DEVICE       offline  HP      Ultrium 6-SCSI
                      /dev/rtape/tape4_BEST    /dev/rtape/tape4_BESTb   /dev/rtape/tape4_BESTn   /dev/rtape/tape4_BESTnb
tape      5  64000/0xfa00/0x19   estape  CLAIMED     DEVICE       offline  HP      Ultrium 6-SCSI
                      /dev/rtape/tape5_BEST    /dev/rtape/tape5_BESTb   /dev/rtape/tape5_BESTn   /dev/rtape/tape5_BESTnb
tape      6  64000/0xfa00/0x1a   estape  CLAIMED     DEVICE       offline  HP      Ultrium 6-SCSI
                      /dev/rtape/tape6_BEST    /dev/rtape/tape6_BESTb   /dev/rtape/tape6_BESTn   /dev/rtape/tape6_BESTnb
tape      7  64000/0xfa00/0x1b   estape  CLAIMED     DEVICE       offline  HP      Ultrium 5-SCSI
                      /dev/rtape/tape7_BEST    /dev/rtape/tape7_BESTb   /dev/rtape/tape7_BESTn   /dev/rtape/tape7_BESTnb

Please note, these drives were not replaced so we have no explanation why they show as "offline".

Also, the drives are "CLAIMED" so executing "scsimgr replace_wwid -D <device file> dsf" does not fix the problem.

Has anyone else encountered this and if so, how did you get the tape devices back "online" without rebooting the HP-UX host?

As always, any assistance would be appreciated.

Thanks in advance.

[Patrick Wallek]

I found this in an HP knowledge base document:

The ioscan -P health command output displayed by ioscan is picking up the health value from the io tree. This health value is updated by estape driver using escsi routines on getting path online/offline notifications from the interface driver.

The health status is marked offline if there are no paths available to the device. This is done on the path change notification from the interface driver to the estape driver. This means that there are no active paths to the tape device at this point. A subsequent open will try out available paths, and then if an active path is present that will be used. An event notification from the interface card to the tape device will trigger the estape driver to mark the health status as online. The tape device should be online most of the times, and will be offline when there is an access problem due to no available paths to the tape device.

In summary, the tape state is changed to offline when the driver loses connection with the tape drive. The health state can be changed to online by accessing (reading from / writing to) the tape drive.

[Bill Hassell]

This has been a problem with tape libraries for years. A power cycle requires accessing all devices a couple of times, then the offline/not ready status goes away. The mc -p /dev...<robot_device> -q command will query the tape library and bring it back online. The command mt -f /dev...<tapedrive> status should bring back each tape drive.