View Single Post
Old 8th August 2008
jggimi's Avatar
jggimi jggimi is offline
More noise than signal
 
Join Date: May 2008
Location: USA
Posts: 7,975
Default

  1. As I mentioned above: if your OS panics, by default the console will enter the ddb debugger. Since this only happens when you're running X, you will never see this happen; your OS will "hang" and you will be unable to reach the console. As I recommended: disable the ddb debugger. Set sysctl ddb.panic to 0. Edit /etc/sysctl.conf appropriately. On reboot, ddb will be disabled. If the OS panics, it will dump memory to swap space, and reboot, and the dump will be saved in /var/crash by savecore(8), as described in the crash(8) man page I recommended.
  2. Increase your swap space. Since we do not know your environment, it is possible -- perhaps likely -- that whatever you are doing is consuming all of your swap space, causing the "hang" you are seeing. In addition, if you want to be able to have the OS save the results of the panic and reboot, you will need to have more swap than you have now. Best practice minimum is 2X RAM size, primarily to manage dumps of RAM. You have very little RAM for modern X applications, and I would recommend 512MB or 1G swap space without batting an eye. FAQ 14.4 describes how to add swap space using a file -- note that doing this adds swap space which is not usable with savecore(8). I recommend adding a 2nd swap partition or repartitioning your drive, as necessary. Also, I do not know if your existing 200MB swap space in a partition will be enough to save 192MB -- depending on existing swap usage or control information it may be insufficient, so if you do not do this, and only set the sysctl, you may still see "hangs" .
Once you've set your OS up to dump memory in the event of a panic, while in X the system will appear to hang during the process while RAM is copied to swap space. The time needed is dependent on the speed of the disk drive to save 192MB of information.

If you make these changes, and the system still hangs, install the "stress" package and run various heavy load tests while in console mode, to see if particular types of hardware stresses cause system failures/hangs.

Last edited by jggimi; 8th August 2008 at 07:27 AM.
Reply With Quote