Messages in this thread |  | | Date | Wed, 10 Feb 2010 16:09:00 -0500 | From | Rik van Riel <> | Subject | Re: Improving OOM killer |
| |
On 02/10/2010 03:54 PM, Lubos Lunak wrote:
> Simply computing the cost of the whole children subtree (or a reasonable > approximation) avoids the need for any magic numbers and gives a much better > representation of how costly the subtree is, since, well, it is the cost > itself.
That assumes you want to kill off that entire tree.
You will not want to do that when a web server or database server runs out of memory, because the goal of the OOM killer is to allow the system to continue to run and be useful. This means keeping the services available...
-- All rights reversed.
|  |