On Wed, 2008-05-28 at 20:36 -0600, Stephen Ward wrote: > Today I was poking around on a CentOS amd64 machine that keeps > randomly crashing. I am not the sysadmin for this machine; he is on > vacation. Anyway, I noticed something that seemed unusual to me. > This machine has 1 GB of RAM, and a grand total of 8 GB of swap > consisting of two 4 GB partitions, each on a separate drive. These > two drives also happen to be in a RAID-1 mdadm device. > > Does a 4 GB swap partition seem excessively large to anyone? I read > the mkswap man page and found the following regarding swap > limitations:
I'll echo the sentiments expressed so far and add a few of my own thoughts. In these days of excessive amounts of RAM it does seem a little odd to have swap partitions at all. Certainly it is possible to build a system without it. But if you run out of RAM completely, the out of memory (OOM) killer takes over and processes start being nuked. That's never a good thing. Have some swap around can save a system. If you're using hibernation then I would recommend going double RAM. Otherwise unless you've got a really RAM starved machine, I wouldn't create more swap than you have physical RAM and I don't think I would create more than 1-2GB total. As far as RAID goes, it's not the worst thing in the world to have swap on a hardware RAID volume, but it's less optimal. You'll get more storage if you leave that one as a raw disk. That may not be a real option so don't sweat it with hardware RAID. With software RAID it's really a no-brainer: don't ever put swap on a software RAID volume. Corey
signature.asc
Description: This is a digitally signed message part
/* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
