From: BlaisorBlade <bla...@ya...> - 2004-06-22 18:44:22
|
Alle 17:14, mercoled=EC 9 giugno 2004, Riccardo Losselli ha scritto: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Hi, it's me again... > I've noticed quite a strange thing.. this is a "free" launched on the hos= t: > > hostxxx:~# free > ~ total used free shared buffers cached > Mem: 2068596 2019692 48904 0 19556 1721472 > - -/+ buffers/cache: 278664 1789932 > Swap: 425592 425592 0 > > If i swapoff -a and then swapon -a > > hostxxxx:~# free > ~ total used free shared buffers cached > Mem: 2068596 2017876 50720 0 10364 1910388 > - -/+ buffers/cache: 97124 1971472 > Swap: 425592 0 425592 > > > > I didn't use the dev/anon patch as i've understood that there's a data > lost risk, correct me if i'm wrong... > The host is running with the ska patch installed and /tmp is > tmpfs /tmp tmpfs defaults,size=3D7168M 0 0 > > Isn't it quite strange that the system uses so much swap space, when in > fact it isn't needed? Well, the really strange thing is another one. There are several reports (e= ven=20 from me) that the SKAS patch for 2.6.x and 2.4.25 onwards leaks kernel=20 memory; if in cat /proc/slabinfo the size-4096 rows shows an increasing=20 number of objects (see man slabinfo about the format), then you have that=20 problem, and maybe the host goes OOM (and it cannot reclaim kernel memory):= =20 so maybe this is the reason of your hangs. But I just checked you used olde= r=20 host, and nobody complained about them (yet). Bye =2D-=20 Paolo Giarrusso, aka Blaisorblade Linux registered user n. 292729 |