[01:59] <wgrant> maxb: poolie saw something strange like that a while ago.
[02:00] <wgrant> maxb: We didn't make the Authorization header connection, though.
[02:00] <wgrant> It was noticed only as a difference between launchpad.net/api and api.launchpad.net... the reason now becomes obvious.
[02:05] <wgrant> Ah, I see you found that bug.
[02:28] <wgrant> I guess squid must be being crap.
[06:29] <lifeless> wgrant: !cite
[06:37] <wgrant> Aha, it worked.
[06:38] <wgrant> I knew that would wake you up.
[06:38] <wgrant> But it seems that requests going through squid are broken, requests not going through squid are not broken.
[06:38] <wgrant> A bit suspicious.
[06:39] <lifeless> or
[06:39] <lifeless> it could be faulty interaction setup in anonymous api requests
[06:40] <wgrant> Doesn't affect staging/dogfood/qastaging/dev.
[06:40] <lifeless> interesting
[06:40] <wgrant> Precisely.
[06:40] <lifeless> so, lets get the squid config
[06:40] <wgrant> That was the plan.
[06:41] <wgrant> NFI what it could be.
[06:41] <lifeless> also we should get squid (tiny cache) in front of qastaging with the analogous config
[06:41] <wgrant> Was going to check appserver logs.
[06:41] <wgrant> But they are big :(
[06:41] <lifeless> gnar
[06:41] <lifeless> 3% packet loss
[06:41] <wgrant> Where?
[06:42] <lifeless> hth am I meant to raid like this
[06:42] <lifeless> my adsl
[06:42] <wgrant> Bah, there's like a hunred appserver logs now.
[06:42] <lifeless> (actually, it may be higher, its 3% on first hop pings; 5% second, 4% on the target server
[06:42] <lifeless> oh -yay-
[06:43] <lifeless> just observed a 20s rtt
[06:43] <wgrant> Hm?
[06:43] <wgrant> Yay
[06:43] <lifeless> not ms
[06:43] <lifeless> s
[06:43] <wgrant> Yeah.
[06:43] <lifeless> 2400ms stddev on ones internet is Not Good
[06:44] <wgrant> :(
[06:47] <lifeless> win
[06:47] <lifeless> 3K
[06:48] <wgrant> Grar devpad is still BST.
[06:51] <wgrant> Ahhh
[06:51] <wgrant> I think I may see.
[06:55] <wgrant> Yeah.
[06:55] <wgrant> Squid puts a port in the hostname.
[06:55] <wgrant> 3128
[06:56] <wgrant> That breaks lazr.restful.
[06:56] <wgrant> Somehow.
[06:59] <wgrant>         url_host_and_http_host_are_identical = (
[06:59] <wgrant>             host == request_host and port == request_port)
[06:59] <wgrant>         if (not url_host_and_http_host_are_identical
[06:59] <wgrant>             or protocol != site_protocol or query is not None
[06:59] <wgrant>             or fragment is not None):
[06:59] <wgrant>             raise NotFound(self, url, self.request)
[07:00] <lifeless> sigh
[07:00] <wgrant> lazr.restful turns the path into a URL by appending it to the service root. Then it checks that the URL matches the request.
[07:00] <wgrant> We should stop squid from inserting the :3128, I guess.
[07:01] <lifeless> I'm just looking for the bug
[07:01] <wgrant> As am I.
[07:01] <wgrant> There we are.
[07:02] <lifeless> u found it?
[07:03] <wgrant> Bug #714038
[07:03] <_mup_> Bug #714038: strange "no such object" error on unauthenticated api calls with parameters referring to other objects <api> <Launchpad itself:Triaged> < https://launchpad.net/bugs/714038 >
[07:03] <lifeless> the squid bug
[07:03] <wgrant> It's a squid bug, not an apache/squid config bug?
[07:03] <lifeless> its a feature IIRC
[07:03] <wgrant> Pardon?
[07:04] <lifeless> exactly
[07:05] <lifeless> https://bugs.launchpad.net/collective.buildout/+bug/438694
[07:05] <_mup_> Bug #438694: squid.conf and Squid 2.6 vs. Squid 2.7 <collective.buildout:New> < https://launchpad.net/bugs/438694 >
[07:06] <wgrant> Hm, so we need to set vport?
[07:07] <lifeless> thats my current wager
[07:09] <lifeless> 2.7s7
[07:09] <lifeless> so yes
[07:09] <lifeless> rt it?
[07:09] <wgrant> Will do.
[07:09] <wgrant> I've been seeing this :3128 crap showing up in OOPSes for a while, had been vaguely wondering what was going on.
[07:10] <lifeless> iz apache bug I suspect
[07:10] <lifeless> information is being lost along the way
[07:10] <lifeless> still, vport=443 is appropriate
[07:10] <lifeless> and should fix
[07:39] <maxb> well, yay :-)
[07:39] <maxb> Always nice to wake up to find a bug has been prepared for squishing :-)
[07:40] <maxb> haha
[07:41] <maxb> Right, so sending Cookie: bananas avoids the bug just as well as sending OAuth :-)
[07:41] <wgrant> Yes.
[22:42] <lifeless> jcsackett: sadly flask is not packaged on Lucid
[22:49] <lifeless> however werkzeug is