[lkcd-devel] About netdump timeout issue -- Help Needed

[jidong x. <jid...@gm...>]

Hi, All,

         Sorry for bother you, because I really need your help
troubleshooting my lkcd issue. Thank you in advance.

         I am focusing on the Linux testing jobs, and recently we need to
setup netdump systems. I referred to lkcd attached Readme file and one
document called lkcd_tutorial , however, I can not achieve my goal, say,
networked based dump. I am really suffered from this situation, so I have to
try this way, i.e. consult to you. What I have done including:

         1. I used two workstations, both machines are setup the OS of
SLES9+SP3, while each machine's architecture is x86_64, and our goal is use
one machine to be the server, and another one is set to be the client. (IP
address of the client is 10.198.90.68 while the server is 10.198.91.32,
their netmask are 255.255.248.0)

         2. On the server machine, I installed lkcdutils and
lkcdutils-netdump-server rpm packages, and on the client machine, I
installed lkcdutils rpm package, I recompiled kernel to enable the support
of LKCD, then modified the configuration files based on our environment.
When the server boots, the netdump-server could start up successfully, and
while the client boots, /sbin/lkcd config command would be run. Here I just
list the result of this command.

client:~ # /sbin/lkcd config



Feb  2 08:58:34 client kernel: Closed dump device

Feb  2 08:58:34 client kernel: dump_netdev: source IP 10.198.90.68

Feb  2 08:58:34 client kernel: dump_netdev: target IP 10.198.91.32

Feb  2 08:58:34 client kernel: dump_netdev: target MAC 00:11:43:e5:d0:ea

Feb  2 08:58:34 client kernel:

Feb  2 08:58:34 client kernel: Network device eth2 successfully configured
for dumping

         3. Then I started up netdump-server services, and I also installed
module of netconsole, with which we can see the client log information on
the server point. However, when I executed the dump on the client, I could
not get the dump files on the server, say, in the directory /var/log/dump,
or /var/spool/crash/, I just can get some log information as below,

client:/usr/src/linux-2.6.5-7.244.lkcd/drivers/net # echo d >
/proc/sysrq-trigger  (run this on the client)

server:~ # tail -f /var/log/messages   (in the same time, run this on the
server)

Feb  5 16:46:18 client SysRq :

Feb  5 16:46:18 client Starting crash dump

Feb  5 16:46:18 client Dumping to network device netdump on CPU 0 ...

Feb  1 21:23:30 server netdump[8077]: Got to many timeouts in handshaking,
ignoring client 0x0ac65a44

------------

In the meantime, I can see such messages on the tty0 of server, these
messages displayed repeatedly.

Now we Start Hello send request

Now we Start Hello send request

Now we Start Hello send request

Now we Start Hello send request

Now we Start Hello send request

Now we Start Hello send request

Now we Start Hello send request

Now we Start Hello send request

Now we Start Hello send request

Now we Start Hello send request

delete outstanding

delete reg

delete dir

delete client

         4. Basically, that's the problem, I just have no idea why the
handshake would always fail, not sure whether you can help us out of here,
thank you very much!



Jason