Re: HD spindown and fs cache writeback
>
> Hi,
> I own a TP310 with Slink and some updates installed on it. It only
> has 16Mb RAM, so disk buffers aren't large and I cannot disable swap
> space. However, I (eventually) got it to let its hard disk alone,
> thus allowing it to spin it down.
> Here is what I did:
>
> > * set noatime on filesystems where you don't care about it anyway
> > * use "mobile-update"
Thank goodness, somebody able to solve it - a statistic of one is waaaay
better than nada.
> Here is what I did not (yet) do because it proved unnecessary for the
> moment:
>
> > * make sure you're not syslogging things you don't care about
> > * use "noflushd"
> > * if there's some process named "sync" murder it and see where
> > its options are found so you can tweak them
> > * Use "hdparm" to advise your disk controller directly that it's
> > okay to sleep at more desirable intervals.
> For the last one, the BIOS setup screen is quite powerful, and sets
> sensible values. No "sync" process over here, I even wonder if I ever saw
> one once in my linux lifetime...
> I do not use noflushd, but mobile-update achieves its work quite very
> well.
Cool... I am strongly getting the feeling that "noflushd" and "mobile-update"
are two different attempts at addressing the exact same feature.
> Additionnaly, I removed 'cron' (I didn't need it) but kept 'atd' (which
> stays quiet anyway), and told VI and others not to keep in sync with
> files, so that the hard disk can spin down during the edition of a file.
You might want to experiment with running things that are usually cron jobs
as at jobs then, e.g. logrotate etc. Alternatively running them out of startup
if you for some reason reboot occasionally.
There's a flavor of cron called hccron which is a patch to cron to adress
running things that happened while you were "off" - hc = home computer -
but probably works for suspend nicely too. What tweaks would cron need to
behave itself better? Seems simple enough, change it back to the has-to-be
HUP'd to-know-its-config-changed, have it "memorize" its tasks.
Seems - sigh
> Unfortunately, I could not find any way to stop the disk activity when a
> program maps in read-write mode a file onto memory (GIMP, netscape cache,
> and so on).
Sounds like a lot of apps to retrain :(
> In general, I would advise to remove from memory any program that uses to
> wake up from times to times, because in many cases is gets paged out
> (thrown to swap-space) in the meantime, and thus has to be paged in each
> time it wakes up.
But you said atd behaves itself... ? Perhaps it is merely small enough to
not swap out?
-* Heather * star@ many places *-
Reply to: