On 25/04/18 01:41, Tobias Wolter wrote: > Cheers, > > On Mon, 2018-04-23 at 22:54 +0200, Tobias Wolter wrote: >> Kudos for the hints, Amos, they're great and'll help me a lot >> (tomorrow, it's 22.54 local here). > > Welp, we haven't really come that far, but what we can tell is that > Squid isn't at fault by itself, as we merely forced it to cache some > bad responses from the Apache. > > For anyone interested: We've narrowed it down to issues accessing files > on the local filesystem or remotely; Apache seems to hang trying to put > the file into an existing HTTP/1.1 keep-alive connection. If you turn > off KeepAlive, it works (delay still present, <10s); if KeepAlive is > on, we can see from tcpdump that Apache will happily respond to > KeepAlive requests, but then dump one of the previous response joined > with another response, without <CR><LF> in between. > > Not sure what's causing the hang yet, but apparently not Squid, so > thanks! Welcome and good luck. FWIW; CRLF is not required between responses. If the first response has a Content-Length the payload ends at the exact byte indicated and the second response starts the byte after - no need for a CRLF. So don't think that absence necessarily a bug in itself. Amos _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
