i'm 99.99% certain that the use of smartctl and/or hddtemp is causing my
system to lose contact with the drives. there's just been far too many
concidental errors of this sort:
hdi: status timeout: status=0xd0 { Busy }
hdi: drive not ready for command
hdi: status error: status=0x58 { DriveReady SeekComplete DataRequest }
hdi: drive not ready for command
at the exact time i've got cron jobs running to do either "smartctl -a" or
"hddtemp", or at a time when i run the command by hand.
at some times the error state is bad enough to cause md to mark the disk
as bad.
at one point i started running hddtemp every 5 minutes for logging
purposes and it took less than 2 days for the system to lose contact with
one of the disks... and this repeated after i rebooted.
system is:
- linux 2.4.19-pre7-ac4
- tyan 2462, dual athlon 1.4GHz (onboard IDE is unused)
- promise ultra 100TX2
- promise ultra 133TX2
- 4x maxtor D740X 80GB (each master on one of the promise channels)
(3x 6L080J4, and 1x 6L080L4)
i've replaced drives (always D740X though), cables, and controllers
(swapping a 100TX2 for the 133TX2 which is in there now).
the problem has appeared on all the IDE ports, so it's not restricted to
any one port/drive. however hdi seems to be particularly sensitive --
even after several controller, disk, and cable combinations. the 4 disks
are in a sw raid5 which is well balanced according to iostat -- except
that hdi is the hottest disk in the box (41C vs 35C 28C 32C).
any suggestions?
obviously i could move to a more recent kernel ... but i stayed back on
pre7-ac4 because it seemed later kernels messed up the promise driver in
some way and i never quite paid attention enough to know what a good
stable 2.4.x ac kernel was post pre7-ac4. suggestions welcome.
any known errata regarding SMART accesses interfering with other
operations?
-dean
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Nov 15 2002 - 22:00:27 EST