On Wed, Aug 24, 2005 at 11:32:14PM +0300, Michael Green wrote:
> LVS freshman here, first post.
> I've setup a 1 director, 4 RS NAT-LVS setup with SLES9 (vendor
> supplied kernel and ipvsadm) and apache 2.0.49. Every RS has access to
> shared /srv/www/htdocs.
> I put a 10M tarball file names snps.tgz into /srv/www/htdocs.
> when I try to download (Firefox/wget) this what I get:
> [michaelg@beastier ~] ->wget http://biolvs.weizmann.ac.il/snps.tgz
> --23:26:17-- http://biolvs.weizmann.ac.il/snps.tgz
> => `snps.tgz.2'
> Resolving biolvs.weizmann.ac.il... 126.96.36.199
> Connecting to biolvs.weizmann.ac.il[188.8.131.52]:80... connected.
> HTTP request sent, awaiting response... 200 OK
> Length: 10,798,002 [application/x-gzip]
> 0% [ ] 0
> 23:26:17 (0.00 B/s) - Connection closed at byte 0. Retrying.
> apache logs show:
> 184.108.40.206 - - [24/Aug/2005:23:26:17 +0300] "GET /snps.tgz
> HTTP/1.0" 200 10798002 "-" "Wget/1.9.1"
> Getting index.php is okay, getting snps.zip is not.
> Any clue on this? I'd equally appreciate if you just point me to
> appropriate reading.
> Director's (host biolvs) configuration:
That is very strange. I assume that apache log is from one of b1-4.
It could be some wird MTU problem, but its probably something more
mundane. Could you send the output of a tcpdump on the linux director,
for both a success (index.php) and fail (snps.zip).
tcpdump -n -e -i any host 220.127.116.11 or icmp
Also, could you show the output of ipvsadm -L -c -n after each test.