Brian's Stupid questions of the day

Tony Hammitt thammitt at kc.rr.com
Fri Sep 15 18:38:05 CDT 2000


IBM's JFS is just starting being ported, but they have actual IBM
engineering support, so it may not take long.  ReiserFS has been being
ported for longer, but they don't have any really huge companies
doing all of the engineering work for them, so it's taking longer.

Technologically, JFS is a very standard high-end UNIX journalling FS.
It journals not only metadata like what files are being changed and
which directorys have updated but also journals the data that is
being changed as well.  They use a separate partition(s) for the
journals, which can be pretty arbitrarily sized.  Usually you want
to have a mirror for the journal so if you lose your journalling
partition's disk, you don't have to fsck.  Other than the journal,
the JFS is extendable but not shrinkable, uses LVM tools to add
space then automatically extends the file system in one step.

ReiserFS is a strange beastie that is based on a binary tree concept
(like directories usually are) but extends the tree concept to how
the data is stored as well.  Integrating this with the new unified
buffer cache cleanly in the kernel is what's taking so long.  The
main advantages here is the ease with which journalling of the
structure can be accomplished and the speed of changing things in
the filesystem.  As far as I know, once RFS is combined with LVM,
we'll be able to grow and shrink filesystems easily.  But ReiserFS
doesn't journal the actual file data, so you may lose data if the
power goes off, you just won't lose any structure information.

All things considered, I think that there will be a place in the
typical user's box for both JFS and ReiserFS.  JFS is slower but
nearly bullet-proof.  Then there's XFS from SGI, which is another
data-journalling FS and ext3 which is ext2 with a metadata journal
file.  So we'll have lots of choices for which one we want to use
soon, but not real soon.

SuSE and Mandrake ship ReiserFS-enabled kernels, but if you want to
compile your own, you have to go get the patches.  None of the other
experimental FS's are to that stage yet.  If you really need a
journalling FS, your only real option is ReiserFS right now.  Kernel
2.5 should have it included along with most of the other ones, but
it will be over a year before 2.5 is stable (educated guess).

Re the second part of the post, I'd check out what your default
domain is, what your /etc/resolv.conf file says and what your routing
table looks like.  Are these boxes hooked together using a hub or
a switch?

Regards,

Tony Hammitt

P.S. Mike:  the default reply-to address isn't set to kclug at kclug.org

Brian Densmore wrote:
> 
> Does anybody know what the difference is between the Journaling File System
> (jfs) and the Reiser File System; and which one I should use?
> 
> Also I have the following setup:
>    1 pc which is acting as gateway and personal desktop - connected via 56k line
>    2 pcs which are gateway clients
>    1 pc which is a Web/e-mail/ftp server.
> 
> >From the webserver I can ping the gateway, but the gateway cannot ping the
> webserver. The hosts entry for the webserver is like this
> 
> 172.200.20.129          www.amason.net
> 
> what is wrong with the entry?
> 
> should it be
> 
> 172.200.20.129          pc1.www.amason.net      www.amason.net
> 
> or what? Also, when I try to ping the other machine using the name or nickname
> it goes out to the internet, and then fails to find them, but punching in the ip
> address works fine!?
> 
> Thanks in advance,
> Brian




More information about the Kclug mailing list