Re: AW: [users@httpd] Apache drops connection but ifdown/up resolves the issue!

"Neil A. Hillard" <hillardn@xxxxxxxxx> · Wed, 01 Feb 2006 13:19:21 +0000

Hi,

Oliver.Schaudt@xxxxxxxxx wrote:

Are you running out of file descriptors with your Apache Process ? [ulimit -a]

I don't believe so, results of ulimit -a:

core file size          (blocks, -c) 0
data seg size           (kbytes, -d) unlimited
file size               (blocks, -f) unlimited
pending signals                 (-i) 32767
max locked memory       (kbytes, -l) 32
max memory size         (kbytes, -m) unlimited
open files                      (-n) 1024
pipe size            (512 bytes, -p) 8
POSIX message queues     (bytes, -q) 819200
stack size              (kbytes, -s) 10240
cpu time               (seconds, -t) unlimited
max user processes              (-u) 32767
virtual memory          (kbytes, -v) unlimited
file locks                      (-x) unlimited

Once this problem happens I can restart Apache, start it with 'httpd 

-X', start it with strace and the exact same thing happens.

As soon as I do a ifconfig bond0.2 down it starts working.  If I then 

bring that interface up it carries on working for a while and then stops!

This box is not in production yet so has virtually no loading and no 

clients (apart from me).  I have the same config running on identical 

hardware that doesn't have a bonded connection to the network or VLAN 

tagging that has been in production for over 6 months and no problems at 

all.

The odd thing is that Apache sees the connection request and then just 

drops it!

Is there any further debugging I can perform to try to locate this?

Many thanks in advance,

				Neil.

-----Ursprüngliche Nachricht-----
Von: Neil A. Hillard [mailto:hillardn@xxxxxxxxx]
Gesendet: Mi 25.01.2006 10:53
An: users@xxxxxxxxxxxxxxxx
Betreff: [users@httpd] Apache drops connection but ifdown/up resolves the issue!

Hi,

	I've got Apache 2.0.55 built from source and am seeing an odd problem 

related to multiple IP addresses on VLANs.

The server has a trunked connection back to our switch and has the IP 

addresses 172.24.12.9, 172.24.11.3 and 172.16.14.104.  These are 

configured as bond0, bond0:2 and bond0.2 respectively.  There are other 

aliases IP addresses on bond0 but they're probably irrelevant for this 

problem.

Apache is listening on all interfaces, port 80 and 443.

I have a client at 172.16.40.32 attempting to connect to port 80 on 

172.24.11.3.  The following occurs:

# telnet 172.24.11.3 80
Trying 172.24.11.3...
Connected to 172.24.11.3 (172.24.11.3).
Escape character is '^]'.
Connection closed by foreign host.

An strace against httpd shows:

[pid 31138] accept(3, {sa_family=AF_INET, sin_port=htons(44894), 

sin_addr=inet_addr("172.16.40.32")}, [16]) = 16

[pid 31138] semop(6914069, 0xb79f0c, 1) = 0
[pid 31138] futex(0x837c474, FUTEX_WAKE, 1 <unfinished ...>
[pid 31130] <... futex resumed> )       = 0
[pid 31138] <... futex resumed> )       = 1
[pid 31138] semop(6914069, 0xb79f12, 1) = 0
[pid 31138] poll( <unfinished ...>
[pid 31130] futex(0x837c43c, FUTEX_WAKE, 1) = 0

[pid 31130] getsockname(16, {sa_family=AF_INET, sin_port=htons(80), 

sin_addr=inet_addr("172.24.11.3")}, [16]) = 0

[pid 31130] fcntl64(16, F_GETFL)        = 0x2 (flags O_RDWR)
[pid 31130] fcntl64(16, F_SETFL, O_RDWR|O_NONBLOCK) = 0

[pid 31130] read(16, 0x83eccf8, 8000)   = -1 EAGAIN (Resource 

temporarily unavailable)

[pid 31130] poll([{fd=16, events=POLLIN, 

revents=POLLIN|POLLERR|POLLHUP}], 1, 300000) = 1

[pid 31130] read(16, 0x83eccf8, 8000)   = -1 ECONNRESET (Connection 

reset by peer)

[pid 31130] gettimeofday({1138182161, 971221}, NULL) = 0

[pid 31130] shutdown(16, 1 /* send */)  = -1 ENOTCONN (Transport 

endpoint is not connected)

[pid 31130] close(16)                   = 0

If I then perform 'ifconfig bond0.2 down', 'ifconfig bond0.2 up' and 

repeat the test everything works as expected!  It carries on working for 

a while and then gives up the ghost again!

Apache is receiving the connection and I can see from a packet capture 

that the three-way handshake is completing but for some reason it's 

dropping the connection - almost like tcpwrappers does!

Whilst Apache is refusing the connections, I can still ssh to the same 

IP address.

This one has me puzzled - if anyone can shed some light on it, it would 

be much appreciated!

Many thanks in advance,

				Neil.

--
Neil Hillard                    hillardn@xxxxxxxxx
Westland Helicopters Ltd.       http://www.whl.co.uk/

Disclaimer: This message does not necessarily reflect the
            views of Westland Helicopters Ltd.

---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx
  "   from the digest: users-digest-unsubscribe@xxxxxxxxxxxxxxxx
For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx