🚀 go-pugleaf

In article <1992Dec16.211411.16924@gn.ecn.purdue.edu> mechalas@gn.ecn.purdue.edu (John P. Mechalas) writes:
>In article <92351.29836.J056600@LMSC5.IS.LMSC.LOCKHEED.COM> J056600@LMSC5.IS.LMSC.LOCKHEED.COM writes:
>>
>>Aye, 'tis true that the "three-fingered salute" in the middle of an app will
>>giving you the "Updating allocation map..." message.  I've also encountered
>>a "volume damaged" (aka "you should have backed up your hard drive, dummy)
>>error on doing so--twice.  After the second time I scrapped Stacker and
>>have had no more bad sectors arise (version 2.0).
>
>>What happens when a program hangs up, though?  You can't exit normally and
>>gracefully.  You have no choice but to reset the machine.  I always tried to
>>avoid needing to turn the machine off while programs were actively running,
>>but it wasn't possible 100% of the time.
>
>I have programs hang all the time on me...and I am constantly rebooting 
>while on a Stacker volume.  As long as the app in question wasn't in the middle
>of a read/write exercise, you are okay.  

I think that John's last sentence is the key: as long as you
don't reboot in the middle of a write (or before the write gets done, if
you're using caching) then you'll probably not do any _serious_ damage.
I presume that if you've updated a file (and completed the update), but
not closed it, then the file is OK, its just the allocation information
that is inconsistent (I seem to recall that some parts of UNIX - I'm
thinking of NFS - only update the file information on close as well).

I suspect that MOST applications don't actually hang up (often) halfway
through a write, and that, because of this, we (collectively) don't
OFTEN see the "corrupted" flavour of failure (or write protection as
stacker does it). When you're writing to a file, then you're in the
commonly executed DOS code (however good that is....) and that it works
most of the time. I also suspect that unless you end up completely
corrupting memory, then even when a program is looping (read "hung"),
then its likely that the timer interrupts that drive cache writes get
executed still, so it probably still works. Maybe we should all be
switching off write caching in any case....?

Neil.

-- 
Neil Watson, ESDG Product Support, Concurrent Computer Corp
G0BLD        227 Bath Road, Slough, Berks, SL1 4AX, England
             Phone: (+44) 753 513360  FAX: (+44) 753 513303 
(nnw@slough.ccur.com, nnw@concurrent.co.uk or 100021,3041 @ Compuserve)