Overview -------- The netdump client package is installed on the machine that is to be monitored. It loads and configures the netdump kernel modules so that the machine on crash will send the oops log and the memory dump to the netdump server. Installation ------------ It can be configured by editing the /etc/sysconfig/netdump file. And it is started and stopped by using the /etc/init.d/netdump init script. Normal usage is to change the NETDUMPADDR line in /etc/sysconfig/netdump to have the ip address of the machine running the netdump server, and then put netdump in the init of the runlevel you normally boot into. You then have to allow netdump ssh key access to the netdump server, by running the command service netdump propagate Be prepared to give the netdump password on the netdump server. This command is doing cat /etc/sysconfig/netdump_id_dsa.pub | \ ssh netdump@$NETDUMPADDR cat '>>' /var/crash/.ssh/authorized_keys2 If you don't want to set a password for the netdump user on the server, you can instead manually do: cat /etc/sysconfig/netdump_id_dsa.pub | \ ssh root@$NETDUMPADDR cat '>>' /var/crash/.ssh/authorized_keys2 Alternatively, you can create local procedures to accomplish the same task, particularly if you do not want to give out the netdump or root password on the netdump server to allow people to set up dumping. You can also set SYSLOGADDR if you want console messages (but not crash dumps, don't worry) sent to a syslog server. Note that for now, the priority is not preserved; all messages go out with priority LOG_NOTICE. You can use netdump and syslog separately or both at the same time; they do not conflict. Testing netdump --------------- The netdump package comes with an example file called crash.c. If this is compiled as a module and loaded into a running kernel it will cause an oops and crash the system. Build the module with gcc -DKERNEL -DMODULE -I/lib/modules/$(uname -r)/build/include -c crash.c and then insert it for an instant oops with insmod ./crash.o