A ProLiant server with Ultrium 6-SCSI tape driven by smartpqi, running Debian 10.13. Server is member of a Windows domain using winbind and nsswitch.conf:
group: files winbind
. Group membership of /dev/nst0
has been set to backup
which is a group provided by winbind. /etc/udev/rules.d/70-tape-persistent-names.rules
has been commented out to prevent Debian to create a local group backup
at boot time (with winbind not yet available) shadowing the winbind group.
This has been working fine most of the time.
Today a user, member of backup
has been working (successfully) with the tape, but upon mt offl
all /dev/nst*
and /dev/st*
devices disappeared and were recreated a couple of seconds later in the (local) group tape
. The log file shows (only):
Jun 21 09:07:29 backup kernel: [15871221.205788] smartpqi 0000:06:00.0: removed 0:0:6:0 0000000000000000 Sequential-Access HP Ultrium 6-SCSI AIO-
Jun 21 09:07:49 backup kernel: [15871240.755657] smartpqi 0000:06:00.0: added 0:0:-:- 0000000000000000 Sequential-Access HP Ultrium 6-SCSI AIO-
Jun 21 09:07:49 backup kernel: [15871240.755816] smartpqi can't handle SMP requests
Jun 21 09:07:49 backup kernel: [15871240.756752] scsi 0:0:7:0: Sequential-Access HP Ultrium 6-SCSI 35GD PQ: 0 ANSI: 6
Jun 21 09:07:49 backup kernel: [15871240.762916] st 0:0:7:0: Attached scsi tape st0
Jun 21 09:07:49 backup kernel: [15871240.762921] st 0:0:7:0: st0: try direct i/o: yes (alignment 4 B)
Jun 21 09:07:49 backup kernel: [15871240.763414] st 0:0:7:0: Attached scsi generic sg0 type 1
This is consistent to the observed behavior, but does not provide any insights (to me). The problem could be reproduces three times and stopped reappearing afterwards. Now loading and unloading works just fine again. winbind
had no connection issues at all during this time, the backup
group was available via getent(1)
.
What could have been the cause of this dis- and reconnects?
And, perhaps more important: Is there any way to pin the group membership of a device to a group which is dynamically provided via winbind? I managed to prevent changes, but as soon as the device is recreated, I still have a problem.