"Jeff V. Merkey" wrote:
>
> On a 4 processor POCA server with 4 hot-swappable 10GB SCSI-II hard disk
> drives that have been configured as a mirror group of 4 mirrored NetWare
> volumes the Adaptec SCSI driver on Linux fails during mirror failover
> testing.
>
> The test is simple and involves pulling active hot-swappable SCSI hard
> disk drives out of the POCA server during heavy file system I/O and
> remirroring operations. The Adaptec SCSI driver immediately disables
> ALL OTHER SCSI devices in the system when it detects an I/O failure on a
> single device.
<sigh>....please check the code in question before proclaiming in ALL CAPS
that it does something that it doesn't. The Adaptec driver doesn't disable
anything, it also does time anything out nor call for any SCSI bus resets.
This is all done at the mid layer of the SCSI code, not at the driver level.
> The SCSI Disks are attached to the SCSI controllers in
> groups of two disks per SCSI BUS. During the failover test, the Adaptec
> driver not only disabled the SCSI bus with the hard drive that had been
> removed, but also disabled the other bus as well, and locked up the
> Asych IO subsystem in Linux.
>
> The Adaptec Driver also kept garbaging the console with "resetting bus
> messages"
Which it does when the mid layer scsi code calls into the aic7xxx_reset()
function.
> repeatedly before completely locking up the server and
> corrupting the EXT2 file system on the active SCSI bus.
Which is to be expected if your systems goes down unclean. How long did you
wait before powering it off?
> On I/O error, or a SCSI command timeout error, the driver should simply
> return I/O errors to the system and leave the remaining SCSI devices
> operational on the system.
To be precise, this is exactly what the aic7xxx driver does, no more, no less.
--Doug Ledford <dledford@redhat.com> http://people.redhat.com/dledford Please check my web site for aic7xxx updates/answers before e-mailing me about problems
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Apr 30 2000 - 21:00:13 EST