> Steve Lord <lord@xxxxxxx> wrote:
> > This adds a new /proc/sys/vm/pagebuf interface which lets you
> > tune a number of pagebuf parameters at runtime. The format is
> > as follows:
>
> > # cat /proc/sys/vm/pagebuf
> > 100 1500 100 128 2048
>
> > The numbers are, in order:
>
> > o sleep interval for the delayed write pagebuf thread - writes out
> > dirty metadata, value in HZ
>
> > o age of delayed write metadata buffers before they are written,
> > value in HZ
>
> > o sleep interval for page cleaner thread, value in HZ
>
> > o maximum cluster size of I/O for delayed allocation conversions,
> > value in pages.
>
> > o maximum delayed allocate pages allowed before page cleaner thread
> > woken up.
>
> are those parameters (and all the other stuff behind /proc/fs/xfs)
> anywhere documented ? - if not it would be a good idea to leave at
> least a basic documentation of then in linux/Documentation/sysctl/
>
> just an idea
You are correct, there needs to be some documentation to go with this,
I am going to be changing how the sysctl interface is presented to be
a one parameter per file interface in /proc.
We have a perl script which interprets the xfs statistics, that should
go into the tree, also the pcp package understands them:
http://oss.sgi.com/projects/pcp/
I still need to get back to you on the power pc stuff, I'm a bit buried
at the moment.
Steve
>
> t
>
> --
> thomas.graichen@xxxxxxxxxxxxx
> technical director innominate AG
> clustering & security networking people
> tel: +49.30.308806-13 fax: -77 http://innominate.de
|