On 20 Apr 1999 11:50:47 -0700  Wolfgang Rupprecht wrote:

> SunOS 4.x had an interesting approach to parity errors.  1) If the
> page was unmodified and backed by some disk it would just get it from
> disk.  2) If it was uncorrectable it would kill only that process.
> 
> I'm not sure 2) was all that much of a wonderful idea since a system
> running on auto-pilot could merrily run with some vital daemon dead.
> A reboot might actually be a better no-brainer.

2) is a good idea if something is logged and the memory stays marked
bad.  'sendmail pid XXX killed due to memory parity error' in a log is
pretty clear about what happened, and for a high availability (-ish)
system is a lot less traumatic than a reboot in the instanace that the
process isn't critical.

Of course, Murphy never sleeps and the process will be some critical
database daemon, or update, or the process monitoring the log file ...

Giles