Dear Amos and Dean, Thank you for your answers. I think, these timeout settings only reduces the failed attempts but doesn't solve the problem. Is there a correct solution for full fail-safe peer selection? In my opinion I do this with ICP queries. Can I do perfect fail-safe peer selection with two proxies between ICP? reverse-proxy <--> proxy with 2 ips <--> web server Best regards, László Király ---------- Original Message ----------- From: Amos Jeffries <squid3@xxxxxxxxxxxxx> To: squid-users@xxxxxxxxxxxxxxx Sent: Thu, 03 Jun 2010 02:10:33 +1200 Subject: Re: fail-safe and load balancing with reverse proxy > Dean Weimer wrote: > > Try using "peer_connect_timeout" You can lower the time out so it fails over faster. > > > > Thanks, > > Dean Weimer > > Network Administrator > > Orscheln Management Co > > > >> -----Original Message----- > >> From: Király László [mailto:kiru@xxxxxxxxxxxxxxxx] > >> Sent: Wednesday, June 02, 2010 3:14 AM > >> To: squid-users@xxxxxxxxxxxxxxx > >> Subject: Re: fail-safe and load balancing with reverse proxy > >> > >> Okay, I compile a brand new 3.1.4 squid with --icmp-enable option. > >> > >> I added also to the squid.conf: > >> --- > >> pinger_program /usr/local/squid/libexec/pinger > >> query_icmp on > >> test_reachability on > >> --- > >> > >> It didn't help. :S > >> > >>> Hi List, > >>> > >>> I use a squid3-3.0.STABLE8 reverse proxy on a debian system. > >>> It makes forward queries to web server, which is accessible from 2 > >>> public ips. > >>> > >>> My peer config: > >>> --- > >>> cache_peer x.y.z.57 parent 80 0 no-query no-digest no-netdb-exchange > >>> originserver name=parent1 round-robin login=PASS weight=16 > >>> cache_peer a.b.c.118 parent 80 0 no-query no-digest no-netdb-exchange > >>> originserver name=parent2 round-robin login=PASS weight=1 > >>> --- > >>> > >>> I would like to do a fail-safe connection to the web server. > >>> > >>> It's working, but if one of the public ips isn't accessible, there > >>> is some Connection timed out (110) proxy message until the parent is > >>> detected as dead, while the proxy tries to query the offline parent. > >>> > >>> How can I eliminate this thing? > >>> Why squid doesn't resend the query to the another parent? > >>> > >>> I cannot set ICP queries while the parent is a simple web server. > >>> Is there a way to make better dead peer detection? > >>> > >>> Can I do this whith icmp queries? > >>> > >>> Best regards, > >>> László Király > >> ------- End of Original Message ------- > > > > 3.1 also brings in the cache_peer options connect-timeout and > connect-fail-limit to set per-peer how long to wait for TCP > connections to failover and how many attempts to allow before > declaring it dead. > > Note: ICMP and NetDB is used to detect the closest source when > multiple are available. Not the live/dead status of a peer. > > Amos > -- > Please be using > Current Stable Squid 2.7.STABLE9 or 3.1.4 ------- End of Original Message -------
