On Wednesday 31 August 2005 04:08, elefino wrote:

Time to fill another gap or two in my knowledge.

In another list, in a thread that I found while googling for info about cyrus and postfix, I ran into a conversation where someone suggested that (for the purposes of the other people in the conversation - most of them setting up or managing fairly large mail systems) ext3fs was not as good as reiserfs, and they should do this:

"In /etc/fstab go to the /var [entry] and change it to something like /dev/cciss/c0d1p1 /var reiserfs data=writeback,noatime 1 2

Reboot the system "

I think the implication was that you can change the fs just by changing an fstab entry. Is that true?

Of course not. You need to reformat the partition to change the file system. The exception being if you move between ext2 and ext3, where all you need to do is create a journal, since that is more or less the only difference between the file systems. But between ext3 and reiser you definitely need a reformat. Also, data=writeback isn't a valid parameter for reiser

I thought that sort of thing would have been laid down at formatting time and just formalized in fstab, and would therefore be a bit more difficult to overhaul than just an edit of one file and a reboot.

Yep. fstab just tells mount what to try when mounting it, it doesn't actually change anything on the disk.

Also, while we're on the subject, is Reiserfs the hands-down better, faster, more reliable-in-both-ordinary-and-extreme-situations, file-system? A Linux Journal article from a couple of years ago seemed to think so, and supplied several feature-comparison tables to prove it. Or is that the kind of question likely to stir up religious wars? (in which case, forget I asked :-)

reiserfs is usually held to be better for file systems with lots of small files, but suffers when you introduce larger files. It's also better at reading than at creating new files, as far as I know. Personally I like xfs as a system with overall good performance, and it's just brilliant for very large files But the general rule is that no one knows what your particular system will do, so for a performance critical server, it pays to examine which file system is best for your particular work load