Help with strange web server problem

Jerry Bell jbell at stelesys.com
Mon Feb 13 18:17:04 PST 2006


Charles - thank you for your excellent investigation!  I'm pretty sure 
that my colo provider isn't running a firewall (I've asked them not to, 
anyhow).  I am running IPFW on that box, with the standard "allow tcp 
from any to any established" followed by the  "allow tcp any to my_ip 80 
setup".  I've done that on other servers without it being a problem like 
this.  I'm going to have the colo double check for router acl's or 
something like that in the morning.

Since this is such an intermittent problem, I can't yet say that it's 
fixed, but I ran with the "disks being idled" theory and wrote a small 
script that creates a file and deletes a file every minute, and since 
that's been running, I've not seeing the issue repeat - but then this is 
not a very repeatable problem.

Thanks again for your great assistance.

Jerry


Charles Swiger wrote:
> On Feb 13, 2006, at 3:12 PM, Jerry Bell wrote:
>> I didn't want to spam the link out, but it's www.musiclodge.com.  I will
>> gather the capture data from working and non working sessions and 
>> send it
>> out.
>
> Well, I can confirm the behavior you've described.
>
> It looks somewhat like a stateful firewall or is in the way and is 
> generating an RST, even while your webserver tries to generate a 
> response.  However, once the firewall sees the outbound traffic, it 
> seems to create a dynamic rule which lets the traffic from subsequent 
> connections through:
>
> 5-pan# tcpdump -tnXs 0 host www.musiclodge.com
> tcpdump: verbose output suppressed, use -v or -vv for full protocol 
> decode
> listening on en0, link-type EN10MB (Ethernet), capture size 65535 bytes
> IP 199.103.21.238.50740 > 63.175.100.44.80: S 2282569549:2282569549(0) 
> win 65535 <mss 1460,nop,wscale 0,nop,nop,timestamp 1159441862 0>
>         0x0000:  4510 003c 4653 4000 4006 7328 c767 15ee  
> E..<FS at .@.s(.g..
>         0x0010:  3faf 642c c634 0050 880d 3f4d 0000 0000  
> ?.d,.4.P..?M....
>         0x0020:  a002 ffff 815f 0000 0204 05b4 0103 0300  
> ....._..........
>         0x0030:  0101 080a 451b adc6 0000 0000            ....E.......
> IP 63.175.100.44.80 > 199.103.21.238.50740: S 2634350592:2634350592(0) 
> ack 2282569550 win 65535
>         0x0000:  4500 0028 0000 4000 2506 d49f 3faf 642c  
> E..(.. at .%...?.d,
>         0x0010:  c767 15ee 0050 c634 9d05 0000 880d 3f4e  
> .g...P.4......?N
>         0x0020:  5012 ffff 03bc 0000 0000 0000 0000 1b60  
> P..............`
>         0x0030:  2678                                     &x
> IP 199.103.21.238.50740 > 63.175.100.44.80: . ack 1 win 65535
>         0x0000:  4510 0028 4655 4000 4006 733a c767 15ee  
> E..(FU at .@.s:.g..
>         0x0010:  3faf 642c c634 0050 880d 3f4e 9d05 0001  
> ?.d,.4.P..?N....
>         0x0020:  5010 ffff 03bd 0000                      P.......
>
> 3-way handshake is completed here, next traffic should be from my 
> machine making the "GET /", request, but instead your machine sends 
> another ACK:
>
> IP 63.175.100.44.80 > 199.103.21.238.50740: S 2238145710:2238145710(0) 
> ack 2282569550 win 65535 <mss 1460,nop,wscale 1,nop,nop,timestamp 
> 1453026167 1159441862>
>         0x0000:  4500 003c 57fa 4000 3206 6f91 3faf 642c  
> E..<W. at .2.o.?.d,
>         0x0010:  c767 15ee 0050 c634 8567 64ae 880d 3f4e  
> .g...P.4.gd...?N
>         0x0020:  a012 ffff 9cdb 0000 0204 05b4 0103 0301  
> ................
>         0x0030:  0101 080a 569b 6b77 451b adc6 9345 1153  
> ....V.kwE....E.S
>
> Interesting that the previous ack had no TCP options set, whereas this 
> one does include a timestamp in response.
>
> IP 199.103.21.238.50740 > 63.175.100.44.80: . ack 396204883 win 65535 
> <nop,nop,timestamp 1159441863 1453026167>
>         0x0000:  4510 0034 4656 4000 4006 732d c767 15ee  
> E..4FV at .@.s-.g..
>         0x0010:  3faf 642c c634 0050 880d 3f4e 9d05 0001  
> ?.d,.4.P..?N....
>         0x0020:  8010 ffff 8157 0000 0101 080a 451b adc7  
> .....W......E...
>         0x0030:  569b 6b77                                V.kw
>
> Where did sequence # 396204883 come from?  And your side follows up 
> with a pair of connection resets, and a normal ACK packet, too.
>
> IP 63.175.100.44.80 > 199.103.21.238.50740: R 2634350593:2634350593(0) 
> win 0
>         0x0000:  4500 0028 b6f6 4000 3206 10a9 3faf 642c  
> E..(.. at .2...?.d,
>         0x0010:  c767 15ee 0050 c634 9d05 0001 0000 0000  
> .g...P.4........
>         0x0020:  5004 0000 cb24 0000 0000 0000 0000 f3fa  
> P....$..........
>         0x0030:  5489                                     T.
> IP 63.175.100.44.80 > 199.103.21.238.50740: R 2634350593:2634350593(0) 
> win 0
>         0x0000:  4500 0028 4bfc 4000 3206 7ba3 3faf 642c  
> E..(K. at .2.{.?.d,
>         0x0010:  c767 15ee 0050 c634 9d05 0001 0000 0000  
> .g...P.4........
>         0x0020:  5004 0000 cb24 0000 0000 0000 0000 abb8  
> P....$..........
>         0x0030:  c9be                                     ..
> IP 63.175.100.44.80 > 199.103.21.238.50740: S 2238145710:2238145710(0) 
> ack 2282569550 win 65535 <mss 1460,nop,wscale 1,nop,nop,timestamp 
> 1453026467 1159441862>
>         0x0000:  4500 003c 3a9d 4000 3206 8cee 3faf 642c  
> E..<:. at .2...?.d,
>         0x0010:  c767 15ee 0050 c634 8567 64ae 880d 3f4e  
> .g...P.4.gd...?N
>         0x0020:  a012 ffff 9baf 0000 0204 05b4 0103 0301  
> ................
>         0x0030:  0101 080a 569b 6ca3 451b adc6 bdd6 d7c9  
> ....V.l.E.......
>
> ...and my side closes, too.  Something is badly confused.
>
> IP 199.103.21.238.50740 > 63.175.100.44.80: R 2282569550:2282569550(0) 
> win 0
>         0x0000:  4500 0028 465a 4000 4006 7345 c767 15ee  
> E..(FZ at .@.sE.g..
>         0x0010:  3faf 642c c634 0050 880d 3f4e 0000 0000  
> ?.d,.4.P..?N....
>         0x0020:  5004 0000 a0cf 0000                      P.......
>
>     -------------------
>
> When I repeat the connection attempt a few seconds later:
>
> IP 199.103.21.238.50743 > 63.175.100.44.80: S 262625798:262625798(0) 
> win 65535 <mss 1460,nop,wscale 0,nop,nop,timestamp 1159442517 0>
>         0x0000:  4510 003c 46c8 4000 4006 72b3 c767 15ee  
> E..<F. at .@.r..g..
>         0x0010:  3faf 642c c637 0050 0fa7 5a06 0000 0000  
> ?.d,.7.P..Z.....
>         0x0020:  a002 ffff 815f 0000 0204 05b4 0103 0300  
> ....._..........
>         0x0030:  0101 080a 451b b055 0000 0000            ....E..U....
> IP 63.175.100.44.80 > 199.103.21.238.50743: S 362624500:362624500(0) 
> ack 262625799 win 65535 <mss 1460,nop,wscale 1,nop,nop,timestamp 
> 1453058903 1159442517>
>         0x0000:  4500 003c e034 4000 3206 e756 3faf 642c  
> E..<.4 at .2..V?.d,
>         0x0010:  c767 15ee 0050 c637 159d 35f4 0fa7 5a07  
> .g...P.7..5...Z.
>         0x0020:  a012 ffff 169b 0000 0204 05b4 0103 0301  
> ................
>         0x0030:  0101 080a 569b eb57 451b b055 55d6 9ceb  
> ....V..WE..UU...
> IP 199.103.21.238.50743 > 63.175.100.44.80: . ack 1 win 65535 
> <nop,nop,timestamp 1159442517 1453058903>
>         0x0000:  4510 0034 46c9 4000 4006 72ba c767 15ee  
> E..4F. at .@.r..g..
>         0x0010:  3faf 642c c637 0050 0fa7 5a07 159d 35f5  
> ?.d,.7.P..Z...5.
>         0x0020:  8010 ffff 8157 0000 0101 080a 451b b055  
> .....W......E..U
>         0x0030:  569b eb57                                V..W
>
> 3-way handshake finishes OK, this time your initial ACK is OK and 
> includes a timestamp back.
>
> IP 199.103.21.238.50743 > 63.175.100.44.80: P 1:17(16) ack 1 win 65535 
> <nop,nop,timestamp 1159442525 1453058903>
>         0x0000:  4510 0044 46cd 4000 4006 72a6 c767 15ee  
> E..DF. at .@.r..g..
>         0x0010:  3faf 642c c637 0050 0fa7 5a07 159d 35f5  
> ?.d,.7.P..Z...5.
>         0x0020:  8018 ffff 8167 0000 0101 080a 451b b05d  
> .....g......E..]
>         0x0030:  569b eb57 4745 5420 2f20 4854 5450 2f31  
> V..WGET./.HTTP/1
>         0x0040:  2e30 0d0a                                .0..
>
> Here was my GET, which you then ACK....
>
> IP 63.175.100.44.80 > 199.103.21.238.50743: . ack 17 win 33304 
> <nop,nop,timestamp 1453059309 1159442525>
>         0x0000:  4500 0034 052a 4000 3206 c269 3faf 642c  
> E..4.*@.2..i?.d,
>         0x0010:  c767 15ee 0050 c637 159d 35f5 0fa7 5a17  
> .g...P.7..5...Z.
>         0x0020:  8010 8218 be99 0000 0101 080a 569b eced  
> ............V...
>         0x0030:  451b b05d db4e 4827                      E..].NH'
> IP 199.103.21.238.50743 > 63.175.100.44.80: P 17:19(2) ack 1 win 65535 
> <nop,nop,timestamp 1159442526 1453059309>
>         0x0000:  4510 0036 46ce 4000 4006 72b3 c767 15ee  
> E..6F. at .@.r..g..
>         0x0010:  3faf 642c c637 0050 0fa7 5a17 159d 35f5  
> ?.d,.7.P..Z...5.
>         0x0020:  8018 ffff 8159 0000 0101 080a 451b b05e  
> .....Y......E..^
>         0x0030:  569b eced 0d0a                           V.....
>
> ...followed by a correct data response.
>
> IP 63.175.100.44.80 > 199.103.21.238.50743: . 1:1449(1448) ack 19 win 
> 33304 <nop,nop,timestamp 1453059329 1159442526>
>         0x0000:  4500 05dc c865 4000 3206 f985 3faf 642c  
> E....e at .2...?.d,
>         0x0010:  c767 15ee 0050 c637 159d 35f5 0fa7 5a19  
> .g...P.7..5...Z.
>         0x0020:  8010 8218 2e5f 0000 0101 080a 569b ed01  
> ....._......V...
>         0x0030:  451b b05e 4854 5450 2f31 2e31 2032 3030  
> E..^HTTP/1.1.200
>         0x0040:  204f 4b0d 0a44 6174 653a 204d 6f6e 2c20  
> .OK..Date:.Mon,.
> [ ... ]
>
> Check your firewall, or see whether your ISP or whoever has put a HTTP 
> reverse proxy in place which is breaking these connections.
>
> ---Chuck
>


More information about the freebsd-questions mailing list