Hello Patrizio, What about removing HAProxy from the mix? The log looks like pound is telling you that it can not connect back to HAProxy:
[Sep 9 12:08:15 bil1] pound: (4332a940) connect_nb: error after getsockopt: Connection refused [Sep 9 12:08:15 bil1] pound: (4332a940) backend 192.168.0.10:80 connect: Connection refused Maybe HAProxy can't connect to a BE web server, so when pound connects back to HAProxy, HAProxy reports "cannot connect" and then pound records "connection refused". I put my $1 on HAProxy being the culprit here. We run a heavy load environment with pound and it never does this. We run it with a Windows stack in the BE and also a CentOS Linux stack, both are solid. -- Jake > -----Original Message----- > From: Patrizio Tassone - Stardata s.r.l. [mailto:[email protected]] > Sent: Friday, September 09, 2011 4:13 AM > To: [email protected] > Subject: [Pound Mailing List] pound crash and emergency config ignored > Importance: High > > HI all, > > we've benchmarking pound for 1 month with high load request and we fix a > memory leak (we send it to the list), but it seems to be ok. > > we put pound in our production environment and now, every random days > (from 1 to 7) we found pound not serving pages and it needs to be > restarted. > > Our infrastructure is pretty simple: pound -> haproxy -> n web server. > > When pound is not responding, we see these information from logfile: > > > Sep 9 12:08:15 bil1 pound: (4332a940) connect_nb: error after > > getsockopt: Connection refused Sep 9 12:08:15 bil1 pound: (4332a940) > > backend 192.168.0.10:80 connect: Connection refused Sep 9 12:08:15 > > bil1 pound: (43d52940) connect_nb: error after getsockopt: Connection > > refused Sep 9 12:08:15 bil1 pound: (43d52940) backend 192.168.0.10:80 > > connect: Connection refused > > it seems to be a problem of the service listening on 192.168.0.10 > (haproxy), but we write a check script that, in case of failure, try a > get from backend to haproxy directly, and it's responding well, so the > problem seems to be pound. > Another external check found a SSL connection timeout, and not a > "service unavaiable" we have to see in case of all backends goes down. > > Another strange issue: we configure pound to use an emergency backup in > case of failure on the primary haproxy: > > > # SSL section > > ListenHTTPS > > Address ssl_static1_mng > > Port 443 > > Cert "/etc/pound/ssl/cert.pem" > > Ciphers > "ALL:!ADH:!EXPORT56:RC4+RSA:HIGH:MEDIUM:!LOW:!SSLv2:+EXP:!eNUL:!EXP-DES- > CBC-SHA:!EXP-RC2-CBC-MD5:!EXP-RC4-MD5:!EXP-DES-CBC-SHA:!EXP-RC2-CBC- > MD5:!EXP-RC4-MD5:!EXP-EDH-RSA-DES-CBC-SHA" > > Client 1500 > > NoHTTPS11 0 > > > > Service > > BackEnd > > Address 192.168.0.10 > > Port 80 > > Timeout 30 > > End > > Emergency > > Address 192.168.0.11 > > Port 80 > > End > > End > > End > > And, watching logs from pound we reported before, i didn't try to > connect to the emergency backend. > > Customers using this service are running all the S.o. and all version of > firefox/IE (from IE6sp2 to the new ones). > > What we can do? We cannot debug in a production environment, trying to > run pound with valngrid is getting slow and slow, and we cannot have a > closed beta-test in this moment. > > Can anyone help us? > > Best Regards, > > patrizio > > > > > -- > Patrizio Tassone Tel. +39 0521 / 46.35.96 > Founder & Editor in Chief Fax +39 0521 / 185.27.46 > Linux&C. - LinuxPratico - H&C. > Founder StarData s.r.l. [email protected] > > -- > To unsubscribe send an email with subject unsubscribe to [email protected]. > Please contact [email protected] for questions. -- To unsubscribe send an email with subject unsubscribe to [email protected]. Please contact [email protected] for questions.
