Hi Michael,

Interesting.. when I do mdadm --stop --scan it stops all the /dev/md* just 
fine, or at least tells me that it does. When I then issue mdadm --assemble 
--scan --config=/tmp/fai/mdadm-from-examine.conf it locks the computer up. My 
ssh session freezes, I can no longer connect to or ping the machine.

I'll make a trip to the data center and see if there is anything on the console 
that offers a clue as to the issue that we're seeing. Perhaps I have a faulty 
piece of hardware or something?

Thanks,

David Dreezer



On Apr 7, 2011, at 2:32 AM, Michael Tautschnig wrote:

> Hi David,
> 
>> I have an interesting issue with one of our servers failing in the partition 
>> task. It's failing on the assemble with an error that the md is already in 
>> use
>> 
>> (CMD) mdadm --assemble --scan --config=/tmp/fai/mdadm-from-examine.conf 1> 
>> /tmp/BLcvisEctM 2> /tmp/gczYIUwN4C
>> Executing: mdadm --assemble --scan --config=/tmp/fai/mdadm-from-examine.conf
>> Command had non-zero exit code
>> (STDERR) mdadm: /dev/md1 is already in use.
> [...]
> 
> I won't have much time to look into this myself before mid next week, so I'd 
> be
> grateful if you could try to figure out the following:
> 
> If an array is running already, will the above assemble-attempt fail in
> precisely this way? I think it should be possible to try the following 
> commands
> after FAI has aborted on this particular server:
> 
> mdadm --stop --scan
> mdadm --assemble --scan --config=/tmp/fai/mdadm-from-examine.conf
> mdadm --assemble --scan --config=/tmp/fai/mdadm-from-examine.conf
> 
> where the first --assemble succeeds, whereas the second one fails. If that's 
> the
> case, we can hopefully fix that by adding the mdadm --stop --scan before doing
> the assemble. (If you like, you can simply hack that in there and if it a try;
> just change the mdadm --assemble ... to  mdadm --stop --scan ; mdadm 
> --assemble
> ...)
> 
> Thanks a lot in advance,
> Michael
> 

Antwort per Email an