Every 3 minutes the time, system load, % used cpu, tcp cons, udp cons,
bindbucket_nr, phy memory, virtual memory, sys cpu%, usr cpu%,
interrupts/sec, and context switches/sec are all droped into one line. Not
only is that stuff sent to a file but sent over the network to the network
logging machine via. logger.
When the system crashed I looked back on that output and viewed the
charts/graphs I make with gnuplot to see what happened.
In a 3 minute period of time the system went from normal VM usage to over
600Mb of virtual memory. The system reached the limit of the swap
partitions and then crashed. You can see how everything was effected and
the exact numbers by going to: http://www.theshell.com/~blyon/crash/
When the crash happened the system took over the cpu and items like ident
and such started to pile up. Ident requests would not finish searching
/proc or whatever they do and more and more would come in so we had a huge
backlog of processes.
>From today on I will be watching for any user process that starts to use
more than normal ammounts of VM, but I don't think a user process did this
due to the ulimits I have set on the system.
The result of this crash is nothing in the log files, and a frozen system.
The only thing the system will do is open TCP connections and sit there.
So if you were to telnet to it, it would open the connection and then do
nothing.
I hope this information helps.
-Barrett
Barrett G. Lyon
Hot Networking DBA TheShell
Use a real computer: www.alphaLinux.org
Hey, do they test this stuff before it's released?
Sure they do... "it compiles, it's ready!"
Barrett G. Lyon
Hot Networking DBA TheShell
Use a real computer: www.alphaLinux.org
Hey, do they test this stuff before it's released?
Sure they do... "it compiles, it's ready!"
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/