>The above combination causes my Sybase server to crash (once every few hours)
>with a 605 error. Anyone seen this before?
>Relevant System Details:
> Solaris 2.4
> Volume Manager version 2.1
> Recommended patches for Sybase.
> Turning off asynchronous writes.
> Moving disks out of volume manager control (This seems to work, but I
> haven't given it enough time).
1. Call Sybase tech support for and ask which Solaris patches are required for running
the server in Solaris 2.4 . I know that we had to load a couple.
2. I know that for a while Online Disk Suite (the volume manager) was not certified
certified for use with the Sybase SQL Server. Make sure that the version that
you are using is O.K. with Sybase.
3. Make absolutely sure that you do not use sector 0 on any disk in your Raid Array.
Here's what happens: Typically, the disk label and other O/S housekeeping
information is written to the first sector on an O/S disk. Sybase, by default,
writes its Object Allocation Map (OAM) to the first available 512K page in a
database. If one of your Sybase virtual devices starts on the first sector, you
run the risk of overwriting your disk label. In the case of logical volume
managers, they also write their housekeeping information at the beginning of the
disk (at least, this is what happens in AIX, I'm not as familiar with Online Disk
Suite or Solstice). Given this scenario, you can have Sybase and the O/S competing
for to write information to the same place. Either way, if you overwrite the disk
label, or the OAM, you are heading for 605 city.
Sybase has some good technical information on this issue. It is a fairly common
problem -- one which we have been through, and have since corrected. Just make
absolutely, positively sure that at Sybase cannot write to the first 2 or 3 sectors on
your disks, and you should be fine. Oh, by the way, we are using asynchronous I/O to
raw UNIX partitions with no problems.
Then again, you may have something else going on...