On Fri, Jan 20, 2006 at 08:57:29PM -0500, Michael P. Soulier wrote: > On 20/01/06 Kevin Kinsey said: > > > I'd almost offer "definitely more likely", as the amount of swap > > shown is less than 1/2 GB. Michael, how much RAM on this box? > > [EMAIL PROTECTED] ~]$ sysctl -a | grep Memory > Virtual Memory: (Total: 720K, Active 537852K) > Real Memory: (Total: 236228K Active 123640K) > Shared Virtual Memory: (Total: 24812K Active: 11880K) > Shared Real Memory: (Total: 12136K Active: 6672K) > Free Memory Pages: 11692K > > Mem: 124M Active, 25M Inact, 75M Wired, 10M Cache, 34M Buf, 768K Free > Swap: 471M Total, 67M Used, 403M Free, 14% Inuse > > It's not a lot, but it's always been fine for me when running Linux as a > headless server. > > > I once set up a box with 256M RAM and 512M swap, which was > > all good && fine until I got a few extra $$ and decided to up the > > RAM to 768M. After that, it died most every evening with Xorg > > running out of swapspace. > > > > If you've more RAM than swap, it's very possible that any process, > > but especially larger ones, could cause problems like this. > > Fortunately, it's not too difficult to add some swap --- provided > > you've some unformatted disk space available. > > I don't think that's the case. I did notice poorer performance while sshing > into the box in the last few days. If the jdk does have a leak, that would > explain it.
While it's theoretically possible, try to assess the likelihood of this being true: the jdk is used by many users, and they don't report memory leaks. On the other hand, your system manifestly has almost no spare capacity to accomodate transient memory spikes. When this happens, you'll easily run your system out of swap. The fact that your system is *already* using 14% of swap (under presumably a "normal" workload) says that it is clearly overloaded: the instant your system starts swapping in order to accomodate its workload, performance will drop in the toilet. Kris
pgpRbUMhORtjV.pgp
Description: PGP signature
