On Thu, 2002-05-23 at 16:45, José Luis Gómez Dans wrote: > Unfortunately, I deleted Alex Hundson's clarification of how > tunefs deals with mounted/unmounted filesystems That'll teach you :-P > One things which didn't occur to me at the time was the size > of the Journal (which in my case is in /.journal, and has a size of > around 4 Megs). Is there an optimal size, or should the smallest size > be used (1024 FS blocks)? You -J'd it? Why? Anyway, as usual, there isn't really an answer. What type of journaling are you using? If you've got up-to-date tools, you'll probably have ordered journally (the default), unless you changed it ;) Since you're only really ordering data to disk and cacheing meta-ops, the size of your journal isn't hugely critical. The journal entries last at most about 5 sec, IIRC, so you need to make sure that the amount of meta-data you're generating is less than 0.75Mb sec-1, otherwise you explode your journal before the kernel comes to write stuff to disk again. With full journalling, that may be a problem, but only really means that the journal is getting flushed more often that usual - I'm not sure what type of performance problem you would encounter with that. FYI: rendevous:/home/hudson_a# ls -lh /.journal -rw------- 1 root root 32M Mar 29 19:09 /.journal rendevous:/home/hudson_a# df -H | head -n 2 Filesystem Size Used Avail Use% Mounted on /dev/sda2 3.9G 2.5G 1.2G 66% / 4G/32M =~ 4 Meg per Gig of space. :) Cheers, Alex.
Attachment:
signature.asc
Description: This is a digitally signed message part