Have you disabled quick_abort (the default setting)?
Disabling quick_abort (i.e. let squid continue to fetch after abort) can
have the unpleasant side effect that a fetch may stall without any easy
way to restart it except waiting for read timeout (default 15 minutes).
If the TCP/IP channel stalls and squid continues to fetch from it, then
any subsequent request for the same object (until Squid times out) will
continue to use this stalled connection.
To enable quick_abort you can set
quick_abort 0 0 0
What squid needs to resolve this, is a different (low) read_timeout for
aborted connections. Another way to resolve this is to use TCP keepalive
socket option, but this increases the network load somewhat, and I don't
know of any way to tune this.
--- Henrik Nordström Arvids wrote: > > Hi! > > I have squid(1.1.3-1.1.7) on linux box(2.0.27),all works > fine,except, that I cant retrieve this URL: > http://cctr.umkc.edu/user/dschmid/graphics/neon.gif > Squid establishes connection, retrieves ~1% and stalls. > If I use another proxy,or without proxy - all OK. > Can you retrieve this URL without problem? > > Thanks,Arvids.Received on Sun Mar 02 1997 - 06:03:29 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:34:36 MST