18 Mar
2005
18 Mar
'05
7:30 p.m.
On Thu, 17 Mar 2005 17:55:12 -0800 (PST), Eric Smith <eric at brouhaha.com> wrote:
True. You wouldn't believe how many times
I've seen other engineers
trying to set the priority of their task higher than those of other
tasks just because they have some vague idea that their task is
"important", rather than any understanding of actual relationships between
the tasks.
Heh... I remember back in the days of VMS 3.x when they gave our
backup operator SETPRI and recoded the backup script to prompt for a
task priority... our backup person quickly figured out that 32 was the
largest acceptable number. Unfortunately, that was hgher than the disk
or tape ACP (ancillary control process)... from what someone else told
me, we had weeks of trashed backups before things were discovered
because the backup process would yank buffers away from the disk
process before they were really full.
Just because one _can_ elevate a process priority does not make it a
good thing to do, especially around high and low limits...
-ethan