Max swap size?

Stuart Jansen sjansen at buscaluz.org
Thu May 29 09:41:20 MDT 2008


On Thu, 2008-05-29 at 06:02 -0600, Stephen Ward wrote:
> Since nobody here has come out and said it, I'm going to assume that 8
> GB of swap isn't likely to be the cause of a random crash on the
> system.

Because 8GB is more than a 32bit system can address, I suppose that it
is conceivable that too large swap could cause a crash. I would expect
instead that the extra space would just be ignored, but maybe you've
found an under-tested corner case.

> Has anybody ever had kernel panics associated with software RAID using
> mdadm?  This server can hum along merrily for days, and then suddenly
> crash, sometimes more than once a day.  As far as I know, the only
> stuff running on it is samba, mysql, litespeed, svnserve, and
> obviously mdadm.

This is an example of why companies should pay for support contracts.
Can you see the panics when the system crashes? Knowing what part of the
kernel was executing during the crash will sometimes point you in the
direction of the culprit.

Such frequent crashes could also be bad hardware. Bad memory, bad IO
controller, etc. Based on the service listed, I'm going to guess that
maybe the server could go offline for  a few hours on the weekend to run
memtest86, vendor hardware diagnostics, etc.




More information about the PLUG mailing list