[squid-users] cache_peer weight in accelerator mode

Hello,

I have a 2.6.STABLE12 box configured in accelerator/reverse proxy
mode. It sends requests to itself and 4 other boxes:

cache_peer x.x.91.140 parent 80 0 no-query originserver
monitorurl=/status.php monitorinterval=15 round-robin login=PASS
name=lwweb1b
cache_peer x.x.91.142 parent 80 0 no-query originserver
monitorurl=/status.php monitorinterval=15 round-robin login=PASS
name=lwweb3
cache_peer x.x.91.145 parent 80 0 no-query originserver
monitorurl=/status.php monitorinterval=15 round-robin login=PASS
name=lwweb4
cache_peer x.x.91.146 parent 80 0 no-query originserver
monitorurl=/status.php monitorinterval=15 round-robin login=PASS
name=lwweb5
cache_peer x.x.91.147 parent 80 0 no-query originserver
monitorurl=/status.php monitorinterval=15 round-robin login=PASS
name=lwweb6

I'm trying to use weight to balance requests so I appended "weight=10"
to each line as a starting point (eventually I'd like to send more
traffic to some servers, less to others). Once that took effect,
almost all the traffic swung to lwweb6. Out of ~2000 requests, only
20 or 30 went to the other 5 servers.

I understand that this was implemented in STABLE5 so I don't see much
explanation of the behaviour. I assumed it would be similar to other
uses of the weight parameter where it's $rtt/$weight, but in this case
the rtt is 0 for all peers.

Have I tripped across a bug or am I going about this the wrong way?

Thanks,

Sean

-- 
Sean Walberg <sean@ertw.com>    http://ertw.com/