[wwwcvs] what I did to steelix

[Wim Vandeputte - Mailinglists only]

hey,

a short explenation of the juggling I'm doing

How cvs works: it's a pig. It creates a lot of temp files in it's chroot tmp
area.

Basicly it's a copy of the whole directory tree structure with it's CVS/
directories and only the files that are in each CVS/ directory
like 

-rw-r--r--   1 wvdputte  wheel    13 Nov  3 10:06 Checkin.prog
-rw-r--r--   1 wvdputte  wheel  8643 Nov  3 10:11 Entries
-rw-r--r--   1 wvdputte  wheel     4 Nov  3 10:06 Repository
-rw-r--r--   1 wvdputte  wheel    26 Nov  3 10:06 Root

This makes the operation very slow, so I generally put it on an mfs:

mfs:29867 on /home/wwwcvs/www/public/anoncvs/tmp type mfs (asynchronous, local, size=912000 512-blocks)

But because the machine has not much memory, I can't make it too big

If you have an extracted directory tree on your end and all the timestamps
on the files are wrong (because you copied the tree with cp without the -p
for example), cvs will duplicate the whole tree in it's chroot /tmp/
directory. Currently our tree is about 121 MB big (due to all the 
presentations).

In that case the /tmp/ directory for one active user grows to:

mfs:29867      426402    118996    307406    28%    /home/wwwcvs/www/public/anoncvs/tmp

So if you get any more errors for disk full, let me know asap

As far as room for the repository itself I have provided some
reserve for the future:

/dev/wd1e     1032654    641492    339530    65%    /home/wwwcvs