Re: Mdadm problem?

On Wednesday August 2, A.J.Dawson@Bradford.ac.uk wrote:

Hi,

I seem to be having a problem with mdadm running on Gentoo.  I recently 
upgraded from 1.12 to 2.5-r1 and then to 2.5.2, with both of the latter 
exhibiting the same behaviour on the machine in question.

The machine is running a RAID1 array and seemed quite happy running mdadm 
1.12.  Running mdadm 2.5.2 however, approximately two days after it is 
booted, a significant amount of disk activity starts.  Unfortunately the 
activity always seems to start overnight while I am not in the office and 
so I've no idea whether it is a sudden start at the high activity level, 
or whether it builds from a low start.  I can hear the disks being 
accessed and the machine's response becomes sluggish.  If I look at the 
activity of the machine using top, mdadm sits at the top of the activity 
table with a low CPU consumption, but high memory consumption (%mem > 
94%).  Note that I also have mdadm --follow --scan & echo $! > 
/var/run/mdadm in my local.start file.

Very odd.  There was a memory leak in 2.5 but that was fixed in 2.5.1
and so should still be fixed in 2.5.2.

Could you try running
 
   ltrace -o /tmp/ltrace -p `cat /var/run/mdadm`

for a while (maybe 5 minutes), then interrupt it and send me
/tmp/ltrace.

That might give some hints as wto what is happening.

Thanks,
NeilBrown