Hi,
OK, ext3 with metadata-only journaling is now running as reliably as I
can make it without making an incompatible change to the journal
format. A usable release absolutely requires a jfs superblock format
change, so I'm releasing 0.0.4c purely as a WIP snapshot before the
superblock changes go in.
ext3-0.0.4c is now on *.kernel.org and on zen2. Look for it in
ftp.*.kernel.org:/pub/linux/kernel/people/sct/ext3/test-DONTUSE/
or ftp.uk.linux.org:/pub/linux/sct/fs/jfs/test-DONTUSE/
This code should work fine under normal running, but IT CAN CORRUPT
DATA on a crash. Anyone wanting to start testing it is welcome to.
Treat this ONLY as a demo snapshot for testing purposes!!!!!
I'm now starting the 0.0.5 branch, which will include the superblock
changes necessary to make recovery over crashes work with the
metadata-only journaling enabled. Converting up will be possible
easily. Converting back from 0.0.5 to a 0.0.3-compatible format will
require unmounting, remounting ext2 and recreating the journal. 0.0.5
will run 0.0.3 filesystems without converting, but only when using
full-journaling, and 0.0.5 will not create 0.0.3-compatible jfs
superblocks.
OK, you've had all the warnings, so if you do want to go ahead and
give 0.0.4c a test, thanks and let me know how it goes.
There has been no performance tuning at all done on this version, but
so far I'm getting speeds of between 75% (running 30-client dbench on
a 200MB box) and 100% (running straight sequential writes) of ext2
write performance on the new code.
Cheers,
Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [EMAIL PROTECTED]
