Tuesday, 21 January 2014

Re: [RFC/CFT] Trusty work for dmraid

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
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=maLe
-----END PGP SIGNATURE-----
On 21.01.2014 16:15, Phillip Susi wrote:
> On 1/21/2014 9:59 AM, Stefan Bader wrote:
>> The udev rule for dmraid does not key on specific raid types but on
>> generic raid usage:
>
>> SUBSYSTEM=="block", ACTION=="add", ENV{ID_TYPE}=="disk",
>> ENV{ID_FS_USAGE}=="raid", KERNEL=="hd[a-z]|sd[a-z]", \
>> RUN+="/sbin/dmraid-activate %k" ENV{DM_STATE}=="ACTIVE",
>> ENV{DM_UUID}=="DMRAID-*", \ RUN+="/sbin/kpartx -a /dev/$kernel"
>
>> Maybe those could get replaced by individual types but removing
>> support completely seemed "simpler" and guaranteed to succeeed.
>
> I was concerned that attempting to activate the array and failing (
> because support had been removed ) would at least cause bogus errors
> in logs, and might trip up the initramfs.

Ah ok, I have not seen that when testing the various stages with the isw raid5 I
have here. It just will not get detected and not brought up.
That reminds me actually that for any real upload, the migration needs to be
made more safe. Not sure whether always pulling in mdadm when installing the
dmraid without isw support would be ok.
But something along this line as otherwise "/" on isw raid would get broken on
upgrade. I have not done this yet. Currently this is just plain implementation
in the tool.