From: Dancer <>
Date: Tue, 04 Nov 1997 12:21:24 +1000

Okay..Now here's a tricky question.

There are times when our parent loses the plot (either the intermediate
networking breaks, or something else goes awry).
When that occurs, we want to go direct until they come back up.
The problem is that we can't rely on ICP to tell us if the parent is
running, because ICP responses get lost when things get busy, and
according to ICP, the parent is down _most_ of the time. By using
single_parent_bypass we query anyway and still get the object...Assuming
the parent is _really_ running.

We want to _never_ go direct if it is...because that costs
24cents/Megabyte if we do. Fetching from our parent only costs 14cents.
I've thought about making a test program that runs off a cron job. Say,
using echoping to fetch a small sample page every 5 minutes or so, and
swap configs to the direct config if the process fails. The squid
reconfigure takes time however. Can anyone think of a more transparent
way of doing things?


Note to evil sorcerers and mad scientists: don't ever, ever summon
demons or rip holes in the fabric of space and time. It's never a good
ICQ UIN: 3225440
Received on Mon Nov 03 1997 - 18:34:21 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:37:26 MST