blob: cd3741db55692865bef88f137ef755189ed21388 [file] [log] [blame]
Subject: ANNOUNCE: mdadm 2.6 - A tool for managing Soft RAID under Linux
I am pleased to announce the availability of
mdadm version 2.6
It is available at the usual places:
http://www.cse.unsw.edu.au/~neilb/source/mdadm/
and
countrycode=xx.
http://www.${countrycode}kernel.org/pub/linux/utils/raid/mdadm/
and via git at
git://neil.brown.name/mdadm
http://neil.brown.name/git?p=mdadm
mdadm is a tool for creating, managing and monitoring
device arrays using the "md" driver in Linux, also
known as Software RAID arrays.
Release 2.6 adds assorted fixes and improvements and a new major mode.
"Incremental Assembly" via -I or --incremental can be used to
assemble an array one device at a time. The idea is that you get
udev to run "mdadm -Iq devicename" on each new block device that it
finds. Anything that is part of an array gets included in an array as
appropriate.
Two special notes:
1/ This is very new code and is probably buggy. It passes a few basic
tests, and helped me find some kernel bugs, but it is still fresh
and should not be considered 'stable'. Please test and provide
feedback.
2/ There is a bug in the linux kernel that makes incremental assembly
not possible in general (you cannot safely remove a drive from an array
that has not yet been started. This is needed if an old device was
detected first). If mdadm detects a kernel which might have the
bug, it rejects --incremental requests.
The bug will hopefully be fixed in 2.6.20 and this mdadm release
contains patches for 2.6.18, 2.6.18.6 and 2.6.19. Apply the
appropriate patch to test --incremental.
Changelog Entries:
- Fixed UUID printing in "--detail --brief" for version1 metadata.
- --update=resync did exactly the wrong thing for version1 metadata.
It caused a resync to not happen, rather than to happen.
- Allow --assemble --force to mark a raid6 clean when it has two
missing devices (which is needed else if won't assemble.
Without this fix it would only assemble if one or zero
missing devices.
- Support --update=devicesize for cases where the underlying device
can change size.
- Default to --auto=yes so the array devices with 'standard' names
get created automatically, as this is almost always what is wanted.
- Give useful message if raid4/5/6 cannot be started because it is
not clean and is also degraded.
- Increase raid456 stripe cache size if needed to --grow the array.
The setting used unfortunately requires intimate knowledge of the
kernel, and it not reset when the reshape finishes.
- Change 'Device Size' to 'Used Dev Size' because it only shows how
much of each device is actually used, not how big they are.
- --wait or -W will wait for resync activity to finish on the given
devices.
- Fix some problems with --update=uuid and add a test.
- If two drives in a raid5 disappear at the same time, then "-Af"
will add them both in rather than just one and forcing the array
to 'clean'. This is slightly safer in some cases.
- Check device is large enough before hot-add: this improves quality
of error message.
- Don't hold md device open for so long in --monitor mode - map_dev
can be slow and interferes with trying to stop the array.
- Support --uuid= with --create to choose your own UUID.
- New major more "--incremental" for incremental assemble of arrays,
intended for use with udev.
Development of mdadm is sponsored by
SUSE Labs, Novell Inc.
NeilBrown 21st December 2006
Blessed Christmas to all.