[...]

> [..]&FAI::push_command( "mdadm -W --stop $md ; sleep 2 ; cat /proc/mdstat", 
> "$pre_deps_cl", "self_cleared_$md");
> 
> > i.e., add the wait, change the sleep to a smaller value (to speed things 
> > up),
> > and print /proc/mdstat? I have absolutely no idea why the kernel claims some
> > device is still busy, and even less so would I know how to figure out what 
> > is
> > still trying to access the device.
> 
> Done. No change.
> 
[...]

Well, at least the output should change, i.e., /proc/mdstat should be included.
Would you mind sending that format.log?

But well, I suppose you are rebooting the system for each run, so we can't have
any old processes hanging around. Then it *must* be the RAID doing stupid
things. May I ask you to perform the following changes?

- comment out the line in /usr/sbin/setup-storage that does the mdadm-startall
  call.
- comment out the push_command("mdadm --assemble --scan ... around lines 293/294
  in Commands.pm.
- replace the mdadm --stop stuff that I had asked you to hack by a simple
  "true".

Would that then work out? If so, it's definitively the RAID that is clobbering
/dev/sdb, but I have no idea why we don't get it released.

Thanks a lot,
Michael

Attachment: pgplpSFDI2nLm.pgp
Description: PGP signature

Antwort per Email an