Henrik Nordstrom wrote:
On Tue, 18 Oct 2005, [ISO-8859-1] Hendrik Voigtländer wrote:
Ah. So if I use exactly the same build with same time out value on all
squid this would explain this behaviour? Should I perhaps reduce
timeout values on the parents?
You could try.
I'll give it a shot.
2 child caches serving users, client distribution bases on network
segments via proxy.pac including failover if one child goes down.
The childs do not know each other.
Ok. In such case a sibling relation between the childs would help hit
ratio.
Is it advisable to use icp/cache digest for the siblings?
An alternative is to change proxy.pac to distribute based on the
destination host rather than the network segment. Drawback is if you use
authentication... (users may be asked twice for their login, once per
child cache..)
Good idea - I have no cache authentication, just a content filter to
keep the filth out of the cache...
Thanks a lot, Hendrik.
