Unwanted reboots - how to fnd the reason?

[drnicolas]

I am facing some strange problems.
A few days before I upgraded to the newest version 1283.

I am not sure, but I think I had at least 2 unwanted reboots. Once I have seen the exact time in the log. The other is only seen by a too-short system time.
So the reason for the reboots is actually unknown, as I did not see any errors/events in the logs.

The NAS is using ZFS RAIDZ and therefore I have 1 big sZFS-dataset share used by CIFS for backups used by BackupExec. Also there is a ZFS-volume which is used by BAckupExec via iSCSI-target. In Windows 2008 this ZFS-volume has a drive-letter assigned.

The iSCSI-based ZFS-volumes is "well filled" (say 1.4T from 1.8T quota is used).

Can this be the reason? What can I do to track the probem ?

[armandh]

the problem of unknown cause rebooting is usually hardware
test the ram
re-seat cards cables RAM etc.
be sure [on-board] un needed functions are turned off in the bios [sound]
are you sure the power is clean.
try running with drives disconnected to rule out drive related problems
replace the bios battery
try a different power supply
a good visual inspection for MOBO defects [bulging caps, cold solder joints.]
boards need not be old to fail
try other hardware entirely

just a word on clean power
in the late 1970s we had installed a digital PBX that would lock up usually over night
eventually tracked to the 7.2KV /120/208 pole transformer with a loose primary connection

[Lee Sharp]

I second that... Start with memtest86+ overnight. That will stress the system fairly well. If it shows bad ram, you are done. If it reboots, try the power supply. Also line conditioners can help.

[drnicolas]

I think it is a problem with the software.

I found that the reboots come when the Backup machine writes to the CIFS. As far as I know at the moment, the ZFS volume exposed as iSCSI is not the problem.

The log do not tell me much - after such a reboot there are only messages written after the boot-process.

Is there a way to KEEP the log entries ? Maybe I can find soemthing in there

[armandh]

it may look like a software problem but if bits of the OS have gone "wonky"
assuming the embedded OS
replace OS, to a new boot device; apply the previously saved config
consumer grade USB flash memory is subject to write wear failure.
the Embedded OS, where the flash is only written to initially and with config changes,
is usually without trouble, but such consumer flash is not suitable for a full install or swap drive.
IMHO consumer USB flash is not to be trusted. all 3 of mine below use IDE industrial flash [very small/cheap SSD]
two are SATA http://www.logicsupply.com/components/s ... h-modules/
one is PATA http://www.logicsupply.com/components/s ... h-modules/
they were loaded from an optical drive with the live CD.
I also like leftover small HDDs that can be set to spin down after the boot of the embedded OS

[armandh]

PS keep in mind that doing something may use more RAM and hit the failed portion, a hardware problem.

[Lee Sharp]

Unstable hardware can work fine with no load, and crash with load. Have you run memtest?

[drnicolas]

In the meantime , I tried memtest with no problem.

I tried to give a more detailed description in te ZFS section.

So far: I can do hundreds of GBs of backups to the iSCSI with no crash. Otherwise I can trigger the crash with backups to the CIFS section of my ZFS