Re: dump device

[Mark H. Wood]

On Thu, 29 Jun 2000 [e-mail] wrote:
[snip]
> Some vendors have the OS do the dump. I agree that there is some risk here.
> I believe the SGI kernel dump does indeed have the paniccing OS do the
> dump.
[snip]
> I cannot comment on what AIX does... I've only been with IBM a short time,
> and have not looked into the crash dump functions.

I don't know what AIX does either.  IIRC, VMS does crashdumps in a
tolerably safe manner by having two entirely separate disk IO subsystems.
One has been tuned relelentlessly for normal use, and has all the bells
and whistles.  The other is small and simple no matter how slow that makes
it.  The small, simple one is used to pull the kernel into memory at boot
time, and to write dumps.

The dumpfile is preallocated and kept open continuously while the kernel
runs, so it can't be moved, resized, or deleted (same as the pagefiles).
There are constraints on the nature of the dumpfile sufficient to ensure
that all of its retrieval information can be cached.  No reads, one short
list of writes, and you are done.

I'm sure that this process must have corrupted at least one disk
somewhere, but I've never seen it happen or heard that it happened
elsewhere.  If you absolutely can't stand the risk, just don't tell the
kernel where to find a dumpfile and you won't get dumps.

-- 
Mark H. Wood, Lead System Programmer   [e-mail]
2000-05-05 13:27:15 GMT -- still no icebergs in the White River