Is maximum swap usage tunable?

bob prohaska fbsd at www.zefox.net
Mon Mar 5 04:57:33 UTC 2018


On Sun, Mar 04, 2018 at 01:40:03PM -0800, bob prohaska wrote:
> make -j4 buildworld 
> 
> was started with the usb-hosted swap partition turned off, leaving only 
> the microSD-hosted swap running.  It got past the first cycle of llvm-tblgen 
> showing only half of the one GB available swap in use, with no errors or 
> warnings. There is some decent chance it'll run to completion, probably
> late tomorrow. For now I'll just see how far it gets. 
> 
The j4 buildworld using only 1 GB swap on microSD really ran out of
swap after a prolonged (half hour or so) interval of swread and pfault
status in top and a stream of 

swap_pager_getswapspace(4): failed
swap_pager_getswapspace(4): failed
swap_pager_getswapspace(3): failed
swap_pager_getswapspace(3): failed
Mar  4 20:02:09 www kernel: pid 25154 (c++), uid 0, was killed: out of swap space
swap_pager_getswapspace(4): failed
swap_pager_getswapspace(4): failed
swap_pager_getswapspace(4): failed
swap_pager_getswapspace(3): failed
swap_pager_getswapspace(2): failed
swap_pager_getswapspace(4): failed
swap_pager_getswapspace(3): failed
swap_pager_getswapspace(4): failed
swap_pager_getswapspace(3): failed
swap_pager_getswapspace(3): failed
swap_pager_getswapspace(3): failed
swap_pager_getswapspace(4): failed
Mar  4 20:02:49 www kernel: pid 25128 (c++), uid 0, was killed: out of swap space
swap_pager_getswapspace(4): failed
swap_pager_getswapspace(3): failed
swap_pager_getswapspace(7): failed

on the console. There were no "indefinite wait" messages at all

It appears that swap on microSD worked much better than
swap on USB, only failing after hitting 85% or more of utilization
and even then not giving up easily. 

The log files for buildworld and gstat are at
http://www.zefox.net/~fbsd/rpi3/swap_issues/
on the chance they're worth looking at. 

bob prohaska







More information about the freebsd-arm mailing list