On Mon, 25 Jan 2010 11:14:34 -0700 David Kreuter <[email protected]> wrote:
> My experience is with the effects of the OOM killer. Perhaps with Linux > on a desktop it is ok as it may pick on non critical processes but in a > virtual machine server environment it represents a drastic out of > storage condition requiring immediate action. I have seen this a few > times on Oracle servers and some emergency relief in the form of > dynamically adding swap helped but eventually the servers needed to be > rebooted with storage adjustments to virtual machine size and Oracle > SGA. It was unpleasant, like stopping a broken dam with a stick of gum. Firstly: you can configure the machine to forbid overcommit. That was a much demanded feature Red Hat added a long time ago and contributed back. On PC servers with ram and disk being cheap that makes a lot of sense for such servers - then the process gets out of memory errors on allocation not OOM (but you'll need more swap 'just in case'). Not sure how it plays out on a mainframe. Secondly: on a modern kernel you can weight processes for killing. Ask your vendor for advice and how much of the stuff is in your system. Alan ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
