On Tue, 31 Jul 2007 12:02:17 +0930 Michael Terrington <michael@terrington.id.au> wrote: > I have an AMD64 host running multiple vservers. One of the vservers > runs a HTTPS reverse proxy (pound) which maps calls through to Apache on > another vserver. sounds like my setup, but with apache2 as reverse proxy: debian etch amd64 host with apache2 (ssl/gzip) forwarding to various other apache2 instances in different vservers (each running a different web application). > Occasionally I get an error where the Apache vserver just doesn't > respond to the reverse proxy for a request. The log on the reverse > proxy indicates "error reading from host". the only time i've seen something like that was with php out of memory (memory_limit) errors that were being passed up from a backend apache2 instance to the reverse proxy and the reverse proxy didn't know how to deal with it. > I've tried tcpdump but I don't see any output which I presume is because > it's internally routed... So how can I debug this further? tcpdump -i lo host <pound> and host <apache> and tcp port 80 > In case this is an old issue, I'm running a pre-built Ubuntu dapper > kernel from http://ubuntu.uni-klu.ac.at/ ... > linux-image-2.6.15-27-amd64-k8_2.6.15-27.45vs2.0.1+2.0.2rc26_amd64.deb for the record i'm running 2.6.21 from debian testing (2.6.18 in etch reports hiccups in my sata drives and degrades my software raid arrays and i haven't upgraded to the recent 2.6.22 in unstable yet). corey -- undefined@pobox.com