At 4:09 PM +0800 4/23/07, Ed Greshko wrote: >Gilboa Davara wrote: >> On Sun, 2007-04-22 at 11:12 -0600, Ashley M. Kirchner wrote: >>> Seems I've asked this before, but I'll be darned if I can find the >>> answer somewhere in my archives. I need to capture a kernel OOPS >>> somehow. I've got a server that seems to be trigger happy and locking >>> up/crashing once every 24 hours or so. The unfortunate thing is that >>> I'm not there when it happens to look on the monitor and see if I can >>> discern anything from there (usually one of the other guys come in >>> before me and they simply hit the reset button because they don't >>> understand the gibberish on screen anyway.) So, I need to figure out a >>> way to capture that OOPS. Since it's not send to syslog, I can't send >>> it to another machine for logging. What other methods are there of >>> doing this? This machine doesn't have a serial link that I can link up >>> to another either. So, anyone? >>> >>> - >> >> Simple. >> - Connect the machine to another machine using a NULL serial console. >> - Add the following parameters to your grub: >> console=ttyS0 console=ttyS0,57600n8 >> (Where ttyS0 is serial0, 57600 is speed and n8 is 8 bits, no parity) >> - Run cu [1, uucp] or minicom and log all the output. > >I think you missed the first post that pointed out the system has no serial >port. :-) > >I think looking into "netdump" and "netdump-server" may be in order. Or possibly the netconsole= kernel boot parameter? Sends data over UDP. -- ____________________________________________________________________ TonyN.:' <mailto:tonynelson@xxxxxxxxxxxxxxxxx> ' <http://www.georgeanelson.com/>
