On Mon, 18 Mar 2013 12:20:57 +0100 Hans-Peter Jansen [off-list ref] wrote:

Am Freitag, 15. März 2013, 23:43:31 schrieb Hans-Peter Jansen:

quoted

Am Mittwoch, 13. März 2013, 11:52:35 schrieben Sie:

quoted

hi,

 wonder if you could try one more test for me.
 With the other (echo BEFORE / echo AFTER etc) tracing still there, change

the /sbin/udevadm settle --timeout=$MDADM_DEVICE_TIMEOUT

 to
 
   /sbin/udevadm --debug settle --timeout=$MDADM_DEVICE_TIMEOUT >
   /dev/kmsg

2>&1

 I found there is a case where "udevadm settle" can exit before the queue
 is

empty.  It seems like a very unlikely scenario, but it seems clear that
something "unlikely" is happening.

 I'm hoping to see
 
    timeout waiting for udev queue
 
 appear in the logs when this runs.
 
 If you could add that and post the 'dmesg' output if you ever get that
 message - or maybe even if you don't - that would be very helpful.

Did that right now and disabled the sleep 10 in front, that provisionally
"solved" this issue BTW. Now waiting for another md issue to occur.

Here we go: (not censored in any way, and I hope to get the "window" right:
md0 was affected this time.

Thanks! Unfortunately it is not as helpful as I hoped, but it does suggest
that "udevadm settle" does sometimes appear to misbehave even if there aren't
any problems with the md array.

Could I ask for one more?
Prefix the "udevadm settle " command with
  strace  -o /tmp/udevadm.trace -s 500

(making sure that you have 'strace' installed) and then post the
"/tmp/udevadm.trace" file.  Hopefully that will at least allow me to rule out
some possibilities.

Thanks,
NeilBrown