heavy problem with raid initialisation

heavy problem with raid initialisation

[Guillaume Rousse]

Hello.

I'm using software raid with mdadm 1.7.0 on a mandrake linux 10.1, but
I'm facing heavy initialisation troubles. The first array /dev/md0 is
automatically created and launched at startup (though mdadm -As in init
scripts), but not the second array /dev/md1.

mdadm --examine --scan --config=partitions creates the second array as
/dev/.tmp.md1, which I can then assemble using an explicit mdadm -A
/dev/sda2 /dev/sdb2 command, but this is unpractical and failproof :/

Here is my /etc/mdadm.conf:
DEVICE /dev/sda* /dev/sdb*
MAILADDR root
ARRAY /dev/md1 level=raid1 num-devices=2
UUID=e5e9302c:1844139b:d30d31e0:6f8477c3
   devices=/dev/sda2,/dev/sdb2
ARRAY /dev/md0 level=raid1 num-devices=2
UUID=99375e60:af2e538e:59f01931:f86b50f4
   devices=/dev/sda1,/dev/sdb1

And here is content of /proc/mdstat once the second array is running:

Personalities : [raid1]
md1 : active raid1 sda2[0] sdb2[1]
      78123968 blocks [2/2] [UU]

md0 : active raid1 sda1[0] sdb1[1]
      78123968 blocks [2/2] [UU]

unused devices: <none>

I've had a look at google, I found some report of similar troubles, some
of them potentially linked with udev. I found some excerpts of mdadm
code that show than this specific /dev/.tmp syntax use is volontary, but
I found no explanation for it.

Any help appreciated.

Re: heavy problem with raid initialisation

[Luca Berra]

On Fri, Feb 10, 2006 at 11:11:24AM +0100, Guillaume Rousse wrote:
>Hello.
>
>I'm using software raid with mdadm 1.7.0 on a mandrake linux 10.1, but
>I'm facing heavy initialisation troubles. The first array /dev/md0 is
>automatically created and launched at startup (though mdadm -As in init
>scripts), but not the second array /dev/md1.
>
>mdadm --examine --scan --config=partitions creates the second array as
>/dev/.tmp.md1, which I can then assemble using an explicit mdadm -A
>/dev/sda2 /dev/sdb2 command, but this is unpractical and failproof :/
>

i think this issue was squashed in a newer version of mdadm
can you try rebuilding the current cooker rpm on 10.1 and try again.
i don't have a 10.1 laying around to test atm.

L.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

Re: heavy problem with raid initialisation

[Guillaume Rousse]

Luca Berra wrote:
> On Fri, Feb 10, 2006 at 11:11:24AM +0100, Guillaume Rousse wrote:
>> Hello.
>>
>> I'm using software raid with mdadm 1.7.0 on a mandrake linux 10.1, but
>> I'm facing heavy initialisation troubles. The first array /dev/md0 is
>> automatically created and launched at startup (though mdadm -As in init
>> scripts), but not the second array /dev/md1.
>>
>> mdadm --examine --scan --config=partitions creates the second array as
>> /dev/.tmp.md1, which I can then assemble using an explicit mdadm -A
>> /dev/sda2 /dev/sdb2 command, but this is unpractical and failproof :/
>>
> 
> i think this issue was squashed in a newer version of mdadm
> can you try rebuilding the current cooker rpm on 10.1 and try again.
> i don't have a 10.1 laying around to test atm.
I updated mdadm to 2.3.1, but the array is still seen as /dev/.tmp.md1.
Putting this instead of /dev/md1 in /etc/mdadm.conf fix the issue, but
it is quite inelegant :/

Re: heavy problem with raid initialisation

[Luca Berra]

On Fri, Feb 10, 2006 at 06:25:30PM +0100, Guillaume Rousse wrote:
>Luca Berra wrote:
>> On Fri, Feb 10, 2006 at 11:11:24AM +0100, Guillaume Rousse wrote:
>>> Hello.
>>>
>>> I'm using software raid with mdadm 1.7.0 on a mandrake linux 10.1, but
>>> I'm facing heavy initialisation troubles. The first array /dev/md0 is
>>> automatically created and launched at startup (though mdadm -As in init
>>> scripts), but not the second array /dev/md1.
>>>
>>> mdadm --examine --scan --config=partitions creates the second array as
>>> /dev/.tmp.md1, which I can then assemble using an explicit mdadm -A
>>> /dev/sda2 /dev/sdb2 command, but this is unpractical and failproof :/
>>>
>> 
>> i think this issue was squashed in a newer version of mdadm
>> can you try rebuilding the current cooker rpm on 10.1 and try again.
>> i don't have a 10.1 laying around to test atm.
>I updated mdadm to 2.3.1, but the array is still seen as /dev/.tmp.md1.
>Putting this instead of /dev/md1 in /etc/mdadm.conf fix the issue, but
>it is quite inelegant :/

mmmm
try adding auto=yes to your array lines in mdadm.conf
or invoke mdadm -As -a

is /dev on udev ?

L.

-- 
Luca Berra -- bluca@comedia.it
        Communication Media & Services S.r.l.
 /"\
 \ /     ASCII RIBBON CAMPAIGN
  X        AGAINST HTML MAIL
 / \

Re: heavy problem with raid initialisation

[Guillaume Rousse]

For sake of list archive completness :)

Luca Berra wrote:
>> I updated mdadm to 2.3.1, but the array is still seen as /dev/.tmp.md1.
>> Putting this instead of /dev/md1 in /etc/mdadm.conf fix the issue, but
>> it is quite inelegant :/
> 
> mmmm
> try adding auto=yes to your array lines in mdadm.conf
> or invoke mdadm -As -a
auto=yes solved the issue, thanks.

> is /dev on udev ?
Yes