formatted space difference: reiser 3.6 vs 4 vs ext3
 

I am seeking an explanation to why I am experiencing a massive 5% difference in available space between reiser 3.6 and 4 (and ext3 for that matter)

Debian Sarge with kernel 2.6.15.4 with reiser4 patch from namesys.com
4x250gb disks in raid5 128kb block size with a total of approx. 750GB or 699GiB

Size reported with df -h

from that I would not hesitate to choose ReiserFS 3.6 for my file server, but I would love to use ReiserFS 4 to get the performance.

After reading bazillon posts about corrupted Reiser and having some unpleasant experience myself I wouldn't recommend it to anyone. XFS - if performance is paramount (must have an UPS), ext3 - good strong filesystem for everybody else.

http://linuxgazette.net/122/piszcz.html

Hmm... I have mostly read about ext3 handling crashes badly and XFS becoming corrupted :S But would still like to know why ext3 is missing 16gb formatted space, that can't be right.

As it is now, Reiserfs 3.6 use the full space while v4 and ext3 does not!

And I do not trust that test 100%, it is made with a 500Mhz P3 and the best filesystems have a fairly large cpu overhead AFAIK.

I do not remember reading XFS becoming corrupted, quite many are complaining it loses data changes made within last (up to 10) minutes in case of sudden power loss. Large CPU overhead is not a plus in my book. In other words, if a filesystem handling causes large overhead it cannot be best. Your lost 16 GB (gb is gram-bit) is probably space allocated to journal data. ext3 with sync transfers turned on is most reliable filesystem out there AFAIK - slower though than async (Linux default). Formatting with -m 0 will lead to fragmentation, but I think you know it.
The choice is yours of course. BTW, Reiser 4 does not support extended attributes, another minus.

I only tried -m 0 to rule out that it could be the source of error :) But still, 16GB journal data, thats a hell lot of data :scratch:

Anyway I did discard XFS due to the fact that I do not have an UPS, but then again, my server is mostly fileserver (that uses /home, all others are using ext3) and not much are written to it. Will it be safe to assume that the data have actually been written after, lets say 5-10 minutes?

I believe so, and it certainly is written when partition is unmounted as it happens during normal shutdown. In general it seems everyone has his/her own opinion on filesystem choice and what I believe is not necessarily best for you. I just wanted to bring your attention to the fact ext3 and XFS both have developed well lately, in particular ext3 has some killer improvements.

Then I will try and make some tests to see if XFS and I will get along :) Does XFS/ext3 support on-the-fly compression of the content? (I believe that ReiserFS do, at least for v4)

Check how much percentage (usually 5%) each filesystem allocates to root to avoid filling it up accidentally.

Useful reading: Filesystem comparison
Next have a look at man chattr. This will give you synopsis what extended attributes are supported.

That have been taken into account ;) but it was not the case here :)

btw Emerson, XFS shows 699GB formatted space as well :)