On Mon, Oct 23, 2006 at 07:44:53PM -0700, Bill Studenmund wrote:
> 
> AIX kinda does this, or used to. In LVM disks, you had a journal per disk 
> or per LVM group. Thus all file systems on a spindle used the same 
> journal.

AIX with JFS has 1+ jfslogs per LVM Volume Group. Each jfslog is shared
by 1+ filesystems. Due to LVM, you can have a filesystem on one physical
disk using a jfslog on another physical disk. The mapping of which
filesystem uses what jfslog is embedded in LVM metadata (VGDA and LVCBs).
When and how this data is sometimes lost is open for discussion. :)
The jfslog appears to the system as just another LV, block special and
raw.

AIX with JFS2 can use the same method as JFS, or can utilize an embedded
"INLINE" jfs2log, which is pre-allocated at filesystem mkfs time, and
appears to be re-sizeable. You can't switch a JFS2 filesystem from
"INLINE" to "OUTLINE" journal.

I have seen situations where the jfslog refuses to replay, due to (I've
assumed) jfslog being on one SAN LUN, filesystem spread over 100+
different SAN LUNs, and relative "ages" of data on each disk getting
out of whack. I've only had an fsck fail once, and that was 10+ years
ago. Oh, not counting various kernel+fs bugs along the way, of course.

Cheers,
-- 
Paul Ripke
Unix/OpenVMS/TSM/DBA
I love deadlines. I like the whooshing sound they make as they fly by.
-- Douglas Adams