pascal tscharner wrote:
> 1999/03/05 17:30:58| Detected DEAD Parent: yom.test.ubs.com/8080/3130
This is when ICP considers the peer dead. Even when digests are
available Squid uses ICP on misses unless you define the peer with the
no-query option.
To make this a bit confusing Squid does not consider the peer all dead
when detecting this. It continues to send ICP request probes and if
there is no other ICP peer alive then even some HTTP requests may be
sent to the peer (the HTTP part is probably a unintentional sideeffect
of the ICP probing).
> 1999/03/05 17:31:18| TCP connection to yom.test.ubs.com/8080 failed
HTTP connection to the peer has failed. Squid tries to use the peer 10
times before considering it completely dead. When the 10'th request has
failed no further HTTP queries will be sent to this peer.
On the first failing request Squid starts a probe to detect when it
comes alive again. This probe starts 30 seconds after the first failing
request and is repeated every 60 seconds until the peer answers. When
this probe succeds in establishing a TCP connection to the peer the HTTP
failure counter mentioned above is reset.
--- Henrik Nordstrom Spare time Squid hackerReceived on Sat Mar 06 1999 - 15:16:45 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:45:09 MST