On Tue, Jul 12, 2011 at 08:35:04PM -0700, Matthew Dillon wrote: > Well, the kernel memory allocator itself doesn't use swap, so it is > more likely a runaway user program (?) and not the kernel. If tmpfs > is being used it could be something filling up the tmpfs mount, including > possibly a process with a descriptor open on an unlinked file. That's > all I can think of.
I don't think this was the case, because I could go to single user mode without a problem, and I could umount /tmp. > Under normal operation idle processes will eventually wind up getting > paged out since there is no sense tying up ram to hold them. e.g. > things like getty and such, but these normally do not amount to very > much use. This is rather subjective, but on a `good' kernel, the swap usage in top's output stays something like this (it has 4Gbytes of physical memory) Swap: 8192M Total, 32M Used, 8192M Free where as on a `bad' kernel, Swap: 8192M Total, 514M Used, 7678M Free, 6% Inuse This looks odd to me, as I've never needed this much swap space even when I was using 6 jails for pbulk. I'm not using jails now. The last known `good' kernel was built from source as of 1e7aaef, but bisecting cannot be automatic because the later kernels are a bit unstable (e.g., sometimes a process locks in objde1 or objde2 and pbulk stalls waiting for it).
