> -----Original Message----- > From: SXB6300 Mailing [mailto:sxb6300@xxxxxxx] > Sent: Monday, August 29, 2005 8:32 AM > To: squid-users@xxxxxxxxxxxxxxx > Subject: Failed to select source > > > Hi everybody, > > 'Coming back with "another" question... I've got a three level proxy > architecture and on the children, I often have the messages "Failed to > select source for ..."(see cache.log below). This is really disturbing > for the users because as an answer they got the error page "Unable to > forward this request at this time. ...". > I previously made some research (I've already posted some messages) on > what could cause the parents being detected DEAD, but so far > I couldn't > find any answer. > The parent isn't overloaded, all and all the network connectivity is > good (speed, duplex, port errors) and they have an average of 70 req/s > during the day (the same for the children but there are 3 > times more of > them). I've checked the firewall logs, no drops or reject. I've also > tried to increase the syn backlog, play with the connect > timeout (squid) > and the fin timeout (/proc), with no success. > > For a child, we always have three parents, and what I find really > strange is that seen from the children (cache.log below), it's always > the three parents that are seen DEAD. I never have the case where only > one parent is seen DEAD ; it's always all the parents that > are seen like > that. > > Any ideas? This really sounds like a network path problem. Something is preventing the child proxy from reaching all three parents at the same time. > At last resort, I'll try to chain both the child and the > parents on the > same network segment, to exclude any firewall problem. Certainly a good thing to check. > > Thx in advance. > > Pierre-Emmanuel > > cache.log : > 2005/08/29 09:11:05| Detected DEAD Parent: parent1/8080/3130 > 2005/08/29 09:11:05| Detected DEAD Parent: parent2/8080/3130 > 2005/08/29 09:11:05| Detected DEAD Parent: parent3/8080/3130 > 2005/08/29 09:11:05| Failed to select source for > 'http://www.espacetoutterrain.com/acces.htm' > 2005/08/29 09:11:05| always_direct = 0 > 2005/08/29 09:11:05| never_direct = 1 > 2005/08/29 09:11:05| timedout = 0 Since you have specified never_direct allow all (or something equivalent), when the parent proxies are not accessible requests fail. Once access to parent proxies is restored, surfing should resume. What are you currently doing to rectify this failure? > 2005/08/29 09:11:05| Failed to select source for > 'http://fr.wrs.yahoo.com/;_ylt=AgTyE2YRxNOBUTyQdclUkDtjAQx.;_y > lu=X3oDMTB > hNjRqazhxBHNlYwNzZWFyY2g-/SIG=11k98fh49/EXP=1125385828/*http%3 > A//fr.sear > ch.yahoo.com/search?p=chats&ei=UTF-8&fr=fp-tab-web-t-1&fl=0&vc > =&x=wrt&me > ta=vl%3D' > 2005/08/29 09:11:05| always_direct = 0 > 2005/08/29 09:11:05| never_direct = 1 > 2005/08/29 09:11:05| timedout = 0 Eliminating anything between the child and parent proxies is a good first step. Chris
