As I expected, there were a variety of responses to my question, which
asked if anyone knew why an IDE disk in an Ultra 10 would crap out when
asked to mirror.
Thanks to the following who responded (there were a couple others which
I deleted in haste -- no offense intended):
Thomas Wardman
Kevin Van Der Hart
Dave Maupin
Aaron Krongard
We isolated our trouble to a disk, the Seagate ST39120. This was
consistent with Aaron's experience.
What drives me nuts is we put these same disks together with standard
issue Sun 9.0 GBs several times before without incident. Only when a
ST39120 and ST39140 were together did we see this problem in SDS. By
the way, we never lost data -- SDS simply insisted on throwing hissy
fits with mirrors and RAID 5's.
We were able to patch up rather than replace the ST39120. The profile
that worked was:
Solaris 7 11/98
106627 (the SDS Jumbo patch)
107834 (pre-requisite to 107458)
107458 (jumbo patch for disk firmware, IDE to SSA)
We got by without the recommended patches, which we would have done if
we had the hour to spare.
After patching, we ripped everything out but the metadb's, and rebuilt.
We had no subsequent problems.
Other notes to responses:
1) There is no shortage of bug reports on IDE. So we want less
expensive Sun workstations -- we didn't *have* to cut corners on disk,
did we?
2) One or two people suggested putting the disks on the same IDE
controller, which worked for them at the expense of performance. The
Ultra 10 hardware guide expressly says that both disks must be mastered
on different controllers. Also see Info Doc 21973 (3/23/00), which says
mirroring only works with two mastered disks. Go figure.
3) Another person said IDE mirroring was not supported at all, which is
no longer true with DiskSuite 4.2. Again, InfoDoc 21973.
Thanks again for all the help. Same day response for free! God bless
the internet.
Regards,
Michael Ernest
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:14:06 CDT