"Mag Gam" <[EMAIL PROTECTED]> writes: > Typically, we create a partition to capture a kernel dump when the system > crashes. Therefore, a system with 16GB of RAM will have a partition with > 16GB.
> How would I scale a system with 64 or 128GB of memory? Any thoughts? As far as I understand, though, kdump doesn't require you to have a massive swapfile; it can be configured to dump other places (though personally, my swap is always at least my ram size for other reasons) http://lse.sourceforge.net/kdump/ most linux folks, though, don't seem to be that into crash dumps; I know when I made the jump from FreeBSD to Linux, this was a big shocker for me, as most of the *BSD shops I've been in, policy was that when something crashed, you didn't trust that box until you figured out why it crashed (usually by GDBing the crash dump and/or banging on the hardware) but the linux folks seem to think that the crash dumps are not so useful. http://lkml.org/lkml/2007/5/25/267 - and this shows in the support for crash dumps. - they seem to change the mechanism for collecting system dumps every year or two. At most of the linux shops I've worked, it's hard enough to get people to get me serial consoles (which we then set up to log, so we at least can tell the difference between the power getting pulled and a MCE) crash dumps are right out. and I can kinda see why; it's something of a pain in the ass under linux. Thing is, it was actually useful under *BSD. But that's my rant, and unquestionably, Mr. Torvalds knows a lot more than I do, so maybe I should just shut up. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
