Squid3.0 provides no option for re-enabling a cache peer

[Andrew]

Binary package hint: squid3

I am using squid with two parent proxies, one being my work proxy and another being one hosted on my home router that I tunnel to using an SSH tunnel. The SSH tunnel is not always active. What happens is that in squid3 on Lucid, it is seeing that my peer is not up (since the SSH tunnel is not up, as I need to use a proxy for my SSH), and it determines that the cache peer is down, and does not check it again.

Since I need squid3 running before I start my tunnel, I have a chicken and the egg problem, squid3 disables my peer before I get the chance to "start" the peer.

I did not have this problem in Karmic, it kept checking the peer and did not consider it down. The "connect-fail-limit=" option is not implemented in squid3.0 (Appears to be new to 3.2) so I cannot use that.

There seems to be no way to force squid to re-enable the peer. Executing "/etc/init.d/squid3 reload" does not fix the problem.

The result is that I always get a "Unable to forward this request at this time." response for the sites going through my home peer, and squid never even tries to contact that peer to even see if it is running (at the time of request).

ProblemType: Bug
DistroRelease: Ubuntu 10.04
Package: squid3 3.0.STABLE19-1
ProcVersionSignature: Ubuntu 2.6.32-22.36-generic-pae 2.6.32.11+drm33.2
Uname: Linux 2.6.32-22-generic-pae i686
Architecture: i386
Date: Tue Jun 8 11:09:00 2010
InstallationMedia: Ubuntu 10.04 LTS "Lucid Lynx" - Release i386 (20100429)
ProcEnviron:
 PATH=(custom, user)
 LANG=en_US.utf8
 SHELL=/bin/bash
SourcePackage: squid3

[Maxime Ritter (airmax)]

I don't think it's a missing feature in the Ubuntu package. If you would like the "connect-fail-limit=" option to become available in the Ubuntu Squid package, you will just need to wait for the Squid 3.2 release (for the moment, still in beta).

That said, there might a bug, which is that squid doesn't see when a cache peer marked as DOWN becomes available again. Because that's what it's supposed to do, and in my experience, did correctly everytime I used it (ubuntu and non-ubuntu systems. But I didn't try with ubuntu 10.04 package). No need for an additionnal option.

Have you tried "/etc/init.d/squid3 restart" ? This should be working. If not, check that the cache peer is really working.

[Amos Jeffries (yadi)]

"connect-fail-limit=" will not help with re-enabling peers. It only determines how many tries are needed to detect when one goes dead.

This behaviour is known to appear when ICP, HTCP and ICMP are all disabled on the peer link. Squid uses them to do background checks for recovered peers. If any one of these protocols is usable the peer will be detected live as soon as one request to it succeeds.
 Squid-3 does not yet have the background HTTP support needed to monitor the peer without causing disruption to the clients.

[Robert Collins (lifeless)]

I thought we had a grace period of 30 seconds or something after which we try again (and if we get through the tcp handshake consider it up). Its possibly worth logging on startup the way in which peers will be assessed.

[Tiago Stürmer Daitx (tdaitx)]

This seems to have been fixed in an earlier squid release.