Stewart Forster writes:
>
>> I've never tried it myself.   But I will confirm that we READ
>> from the outgoing socket.  But we never WRITE to the incmoing
>> socket.
>
>	By never WRITE, you mean that ICP queries are never sent out the
>incoming socket (unless outgoing == incoming of course)?  Replies to a
>query sent to our cache looks to me to be sent back on the incoming socket
>since icpHandleIcpV2 just writes back onto the socket that was passed to
>it (if it got a query).
Yes, it looks like you are correct.
But I found something strange too.  In squid-1.1 we always queue ICP
replies.  There is only one queue.  So it seems like all ICP replies
will be sent out on the last socket to recieve a query.  It is possible
that we received a group of queries on both sockets but the the replies
would all go out on one socket or the other.
This is much less likely to happen with squid-1.2 because we always try
send() before queueing.   I have never seen ICP replies queued on any
of my caches:
Connection information for squid:
        Number of ICP messages received:        7706371
        Number of ICP messages sent:    7803266
        Number of queued ICP replies:   0
Duane W.
Received on Tue Jul 29 2003 - 13:15:50 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:11:47 MST