Your message dated Sun, 20 Aug 2006 17:02:13 -0700
with message-id <[EMAIL PROTECTED]>
and subject line Bug#381057: fixed in mdadm 2.5.3.git200608202239-1
has caused the attached Bug report to be marked as done.
This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.
(NB: If you are a system administrator and have no idea what I am
talking about this indicates a serious mail system misconfiguration
somewhere. Please contact me immediately.)
Debian bug tracking system administrator
(administrator, Debian Bugs database)
--- Begin Message ---
Package: mdadm
Version: 2.5.2-9
Severity: wishlist
Documenting this possible future change, which came up during the
discussion about mdassemble, which is lighter than mdadm and could
thus be used during boot, or in the udeb.
----- Forwarded message from Neil Brown <[EMAIL PROTECTED]> -----
> I had another question about mdassemble... it requires mdadm.conf,
> right? I currently just create that file during boot because of
> http://bugs.debian.org/284022 -- is there a better way you suggest?
> I could just copy the file into the initramfs from the host, but
> then maybe there have been changes made to the RAID configuration
> without /etc/mdadm/mdadm.conf being updated, and noone noticed
> because everything continued to work because of the initramfs...
Yes, mdassemble and mdadm --assemble require mdadm.conf.
It is needed to identify the arrays to be assembled.
This may seem to make it difficult to just "Assemble all arrays that
you can find", but that is very deliberate. I think doing that is the
wrong thing to do. It can get confused when moving drives from one
machine to another.
To address this issue I wrote the 'homehost' stuff.
Check the man page for details.
The idea is that all arrays belonging to a particular host are tagged
with the host's name, and then they can be found and assembled
reliably.
So:
in mdadm.conf put
CREATE homehost=<system>
(literally) and every array created with be tagged with the hostname
from gethostname(2).
In initrd, put
mdadm --assemble --scan --homehost='<system>' --auto=yes
--auto-update-homehost
and make sure the hostname is set at this point (that might be the
tricky bit).
Then mdadm will scan all devices for md array components that are
tagged for this host and will assemble them all (without a config
file).
If it doesn't find any arrays, then the --auto-update-homehost thing
comes into play.
It will rescan all devices looking for any md - nomatter what host
it is tagged for - and will rewrite the superblocks to be for 'this'
host and will assemble them.
On subsequent boots it will find some arrays tagged for this host
and will assemble them and so will NOT do the auto-update thing.
That only happens when NO md arrays for this host are found.
The net result should be pretty much what everyone seems to want, but
with enough caution to make me happy.
The tricky bit, as I said, is making sure the host name is available
to the initramfs.
NeilBrown
----- End forwarded message -----
Making the hostname available to the initramfs is as easy as adding
a line to the hook.
--
martin; (greetings from the heart of the sun.)
\____ echo mailto: !#^."<*>"|tr "<*> mailto:"; [EMAIL PROTECTED]
spamtraps: [EMAIL PROTECTED]
... and don't get caught in the .NET!
signature.asc
Description: Digital signature (GPG/PGP)
--- End Message ---
--- Begin Message ---
Source: mdadm
Source-Version: 2.5.3.git200608202239-1
We believe that the bug you reported is fixed in the latest version of
mdadm, which is due to be installed in the Debian FTP archive:
mdadm-udeb_2.5.3.git200608202239-1_i386.udeb
to pool/main/m/mdadm/mdadm-udeb_2.5.3.git200608202239-1_i386.udeb
mdadm_2.5.3.git200608202239-1.diff.gz
to pool/main/m/mdadm/mdadm_2.5.3.git200608202239-1.diff.gz
mdadm_2.5.3.git200608202239-1.dsc
to pool/main/m/mdadm/mdadm_2.5.3.git200608202239-1.dsc
mdadm_2.5.3.git200608202239-1_i386.deb
to pool/main/m/mdadm/mdadm_2.5.3.git200608202239-1_i386.deb
mdadm_2.5.3.git200608202239.orig.tar.gz
to pool/main/m/mdadm/mdadm_2.5.3.git200608202239.orig.tar.gz
A summary of the changes between this version and the previous one is
attached.
Thank you for reporting the bug, which will now be closed. If you
have further comments please address them to [EMAIL PROTECTED],
and the maintainer will reopen the bug report if appropriate.
Debian distribution maintenance software
pp.
martin f. krafft <[EMAIL PROTECTED]> (supplier of updated mdadm package)
(This message was generated automatically at their request; if you
believe that there is a problem with it please contact the archive
administrators by mailing [EMAIL PROTECTED])
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Format: 1.7
Date: Mon, 21 Aug 2006 00:25:22 +0100
Source: mdadm
Binary: mdadm mdadm-udeb
Architecture: source i386
Version: 2.5.3.git200608202239-1
Distribution: unstable
Urgency: low
Maintainer: Debian mdadm maintainers <[EMAIL PROTECTED]>
Changed-By: martin f. krafft <[EMAIL PROTECTED]>
Description:
mdadm - tool to administer Linux MD arrays (software RAID)
mdadm-udeb - tool to administer Linux MD arrays (software RAID) (udeb)
Closes: 381057 381303 382263 382389 382450 382480 382876 383688 383806 383891
Changes:
mdadm (2.5.3.git200608202239-1) unstable; urgency=low
.
* Tracking upstream git releases.
- now the --run switch behaves as stated in the manpage. This properly
fixes #287415
- new version-1 partitionable arrays are now named X instead of _dX (e.g.
0 instead of _d0) for device names like md_dX. I actually think this is
a bug and hope upstream will use dX (e.g. d0) instead in a future
version.
- we specify --symlink=no and thus disable the new feature to create
/dev/mdX symlinks to /dev/md/X devices until the entire device node mess
is cleared up. No need to introduce yet another complicating factor at
this stage.
* Instead of trying to do a whole lot of magic with respect to detecting
RAID devices to start, mdadm from now on requires a valid mdadm.conf file
to be installed. It still tries to do what it can, but there are no
heuristics anymore. See /usr/share/doc/mdadm/README.upgrading-2.5.3 .
* Removed the zero-superblock warning because we require the user to sign
off the configuration file anyway.
* This also enables us to use mdadm.conf from the initial ramdisk and thus
closes: #381303.
* In case the user chooses to assemble all arrays from the initial ramdisk,
use the new homehost feature of mdadm (closes: #381057). This will start
only those arrays belonging to the local system, unless it is the first
run on a system, in which case it will start all arrays and mark them as
belonging to the local system.
* Improved the debconf control script: integrated error messages into the
frontends, and made it a bit smarter. Error messages can now be
translated, and the script checks whether the user's choice is listed in
the configuration file and only proceeds if it is, or the user chooses to
ignore that it is not.
* mkconf can now take a generate/force-generate parameter to write directly
to /etc/mdadm/mdadm.conf. A second parameter specifies an alternate
filename.
* mkconf now outputs comments for the settings it suggests.
* Removed all udev-related stuff. We must coexist with udev because there
are setups that assemble arrays without mdadm, so the device nodes must be
created by udev, if that is used.
(closes: #382263, #382450, #383688, #383891, #383806, #382480)
Staying at low urgency since these (RC) bugs only exist in unstable.
* Now installs MD modules and mdadm/mdrun into initial ramdisk regardless of
whether the hook script thinks there are devices to start. This was done
to enable recovery from the initramfs shell.
* Now uses 'MD' instead of 'RAID' consistently in all messages.
* Now rebuilds initramfs for all installed kernels.
* Now breaks the endless config loop only when the user does not see the
question (see #381284, #381007).
* Don't fail mdadm-raid when /dev is on a read-only filesystem
(closes: #382876).
* Updated debconf translations:
- French by Florentin Duneau (closes: #382389).
Files:
4b7ff47f67a0cb0b3e304de701a95d5d 760 admin optional
mdadm_2.5.3.git200608202239-1.dsc
8006fce30bdac088dfb396410428377b 176469 admin optional
mdadm_2.5.3.git200608202239.orig.tar.gz
11e33373fcaf6439cf70cf4c25d6a7d2 84117 admin optional
mdadm_2.5.3.git200608202239-1.diff.gz
40fa2bfdc60770eb1f0e58105bcd6b24 194126 admin optional
mdadm_2.5.3.git200608202239-1_i386.deb
1017b063b913f43dbe7e8f23dd427d03 68188 debian-installer optional
mdadm-udeb_2.5.3.git200608202239-1_i386.udeb
Package-Type: udeb
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)
iD8DBQFE6PP5IgvIgzMMSnURAndkAJ0a8BYhN+jfjkGuGtzG/phkZDfczgCgwjvg
PhYHYrDEZtB7Y4lLtyLp6Bk=
=FDVN
-----END PGP SIGNATURE-----
--- End Message ---
