On Fri, Nov 5, 2010 at 08:45, Iain Buchanan <iai...@netspace.net.au> wrote: > OK so vm.swappiness seemed to help a bit but today I notice that swap > usage is up again. It's firefox: > > PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > 14072 iain 20 0 1369m 897m 15m S 3 29.5 113:14.91 firefox > > I think that's 1.3Gb + 900Mb... sounds like a memory leak to me. > > Anyone else run firefox for 113+ hours? I'm using 3.6.9-r1. >
1.3G is the grant total of Res and Swap. You need to read man top before judging not-entirely-accurate values reported by top. 900M is resident on your main memory. '113+ hours' is not a decent information to draw conclusion from. Running firefox for 113+ hours with a single tab on a text-only website is not same as running dozens of tabs with dozens of multimedia/embedded objects. You say swap usage was up again but dont give data about it. 472M is what top would report in the above case but it is an unrealictic values as it is uncompressed size. output of cat /proc/swaps or free is what one needs to see. Also make sure that you put the swapiness value which you say worked for you in /etc/sysctl.conf. -- Fatih