>
>On 16 Nov 1996 [EMAIL PROTECTED] wrote:
>
>> 0./ Does anyone use Sparc Storage arrays. If so, what management
>> software do you use, the storage Volume Manager s/w (I have version
>> 2.1.1) or Online: DiskSuite?
>>
>
>We tested Volume Manager once, but ran into the problem that it has its own
>fsck. The consequence was that after a crash we lost all AFS files. And with
>the AFS fsck nothing worked. This was more or less in a hurry so I don't know
>whether one could have come around the problem.
We've been running Sun Arrays using Volume Manger for several years
now and have never found that Volume Manager has a fsck which is
"different" from the one provided with the OS. (Solaris 2.3, 2.4, 2.5, 2.5.1).
The Transarc version of fsck did have some problems with the
strings ( e.g. /dev/vx/dsk/3377dg/vicepa ) but those have been fixed
well over a year ago.
A word of caution...make sure that the server not only starts up but
that it also salvages correctly. ;-)
>
>DiskSuite however works - as it would on normal disks. I have heard that
>logged filesystems were problematic, but raid (for which we use it) works
>fine. We haven't tested mirrors. Anyway, you don't need DiskSuite to
>access the StorageArray disks.
>
>
>
>> 2./ How do you configure your disks? What RAID level? I was thinking of
>> either 5 or 0+1 for a pair of internal disks with the OS on them, and
>> RAID 0 for the disks in the storage array for the afs partitions.
>> Exactly how what use do people make of the volume Manager? (the
>> configuration options seem endless).
We typically use RAID 0 or 0+1, we have tested RAID 5 but it's much
slower on the Sun SSA's because it's software based. We have recently
tested MTI's hardware based RAID 5 and found performance to be much
better than Sun's. I've heard a rumor that Sun will soon be reselling
a hardware based controller which could be swapped out with their
Model 210 array controller (nothing for the Model 100 series though)
which would increase performance for RAID 5. Our tests on the MTI array
showed equivalent performance for their RAID 5 and Sun's RAID 0.
Rich
Rich Sudlow
Office of Information Technology
University of Notre Dame