Date: Tue, 05 Apr 2011 09:44:34 -0700 From: Chuck Swiger <cswiger@mac.com> To: Pete French <petefrench@ingresso.co.uk> Cc: FreeBSD-STABLE Mailing List <freebsd-stable@FreeBSD.org> Subject: Re: Kernel memory leak in 8.2-PRERELEASE? Message-ID: <62BEFA1C-9A7D-4981-9F71-6DE15973B7C1@mac.com> In-Reply-To: <E1Q77eY-0001Mc-Ke@dilbert.ticketswitch.com> References: <E1Q77eY-0001Mc-Ke@dilbert.ticketswitch.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 5, 2011, at 7:54 AM, Pete French wrote: >> Having swap provides some cushion. Swap kind of smooths any bursts. (And it can >> also slow things down as a side effect) > > This is why I got rid of it - my application is a lot of CGI scripts. The > overload condition is that we run out of memory - and we run *way* out > of memory .... its never just a little overflow, it;s either handleable or > completely crushed. But swap makes that mre llikely to happen, because > as the processes are swapped out they run slower, take longer to > finish and thus use memory for longer. > > What I saw was that as soon as any web server would start tos wap it would > swftly fall down. Without swap they stay up, but reject requests. Its a better > failure mode... You'll be better off providing swap space with every machine, and tuning Apache's MaxClients to a suitable value such that you don't swap excessively under load spikes. Without any swap, the system will be unable to page out unused portions of processes and will handle less load than it would with swap available. Regards, -- -Chuck
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?62BEFA1C-9A7D-4981-9F71-6DE15973B7C1>