> On Mon, 5 Mar 2001, Joshua Baker-LePain wrote:
>
> > Which, finally, brings me to my question. What is the current
> > conventional wisdom on the best way to get xfs up and running? I'll
> > obviously be testing before I put this into production. Should I be
> > grabbing the latest CVS snapshots, or using the 0.9 pre-release? I'll
>
> Probably you should begin with the pre-release. Do your testing, and then
> try current CVS. Do some more testing. My experience has been very good
> with CVS (once I got it compiled - thanks guys!) but I don't use it in a
> big, high demand, mission critical environment (unless you call my
> slackware mirror and mp3 collection mission critical :) ).
>
> James Rich
> james.rich@xxxxxxxxxxxxx
I would say that the beta images (with some provisos about mouse configuration
due to devfs) is the easiest way to get a system up from scratch. I do NFS
based installs in about 20 minutes - grab all of redhats updated iso images
as well though. The first thing I do after this is build by own kernel with
the drivers I use built into the kernel and devfs turned off, this gets rid
of the need for an initial ram disk and usually speeds the boot process.
The CVS tree should be more stable for you, and offers more features
[2.4.2 base, better VM integration of delayed write, O_SYNC support,
O_DIRECT support, faster shutdown], and probably better performance than
the beta images do. Upgrading the kernel with the cvs tree version after
installing the beta should be a good thing to do. In general we are trying
not to break the cvs tree now, but it is always good to keep one kernel
you know works (the beta images for example) in your lilo configuration.
Steve
|