Date: Tue, 6 Feb 1996 10:40:27 -0500 (EST) From: jlrobins@zappa.cs.uncc.edu (James Robinson) To: Duncan.Barclay@pa-consulting.com (Duncan Barclay) Cc: freebsd-hackers@freebsd.org Subject: Re: Swap adding daemon Message-ID: <9602061540.AA13100@zappa.cs.uncc.edu> In-Reply-To: <31178897@SMTPGATE.PA-CONSULTING.COM> from "Duncan Barclay" at Feb 6, 96 08:47:00 am
next in thread | previous in thread | raw e-mail | index | archive | help
> I am going to write a daemon to keep an eye on the ammount of swap used > and add extra to the system using the vn driver. if a threshold is crossed. > I ensivige this to be useful to people with home machines which are rebooted > most days and dont always need loads of swap space. The swap files > created would be deleted on the next reboot by rc or as part of the daemon > configuration. Good idea! And if the ability comes to unswap from a device, then it could do that too! Sortof of an amd for swap. > > Questions: > Has anyone done it before. > > Would a fstab like config script be better than > automatically finding spare vnodes in /dev and > allocating space. > ie. > #file vnode size threshold > # to create on > /usr/tmp/swapfile1 /dev/vn0a 12M 80% > /usr/tmp/swapfile2 /dev/vn0b 12M 80% > /disk2/tmp/swapfile /dev/vn0c 24M 90% > where the threshold is the percentage of currently used swap. Hmm. fstab syntax is the way to go. I assume that the system, in this case, would add /usr/tmp/swapfile1, after creating it to be 12M, first, once the system exceeds 80% swap capacity. Then, after some time, swapfile2 is joined in after the previously augmented swap space exceeds 80%. Etc. > > Should the swap usage be looked at over a long time to determine > threshold or be faily reactive? Hmm, since we can't remove swap at this moment in time, I'd lean towards the more conversative. What two threshold fields, one to specify 5 minute avg swap usage, and the other to specify one minute swap usage? So, it would add that space if the one minute swap load reading exceeded 95%, or add it if the 5 min swap load average exceeded 75%. This would handle a heavy swap hog (netscape, anyone?), or a general increase in load over time. Just my $.02 -- good idea! Heck, you could prototype it in a script language.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9602061540.AA13100>