On 28/10/15 11:08, Nathan Scott wrote:
Hi Ken,
----- Original Message -----
I'm seeing different failures across a number of machines (*.out.bad files
attached).
There is a common theme, the observed and expected values are off by 2 ...
which smells like a "block" to 1Kbyte conversion issue.
But there are lots of passes as well ... perhaps if there is _no_ activity
then zero is zero, even if the multiplier is wrong?
Another interesting point from looking over those bad files is that it is
only ever the first part that fails - the extended I/O stats are always fine.
For one of those failing cases, could you send through the .full file also?
I'll keep digging into it once I have that, thanks (no failures here FWLIW,
anything interesting about those VMs btw? 32 bit only, anything like that?).
.full files attached.
Machines are ...
vm07 3.10.8 x86_64 Debian 6.0.10
vm11 3.10.8 i686 Debian 6.0.10
vm14 3.10.8 x86_64 CentOS6.6 (Final)
vm28 3.10.8 x86_64 RHEL Server 6.7 (Santiago)
so no obvious pattern there.
vm07-975.full
Description: Text document
vm11-975.full
Description: Text document
vm14-975.full
Description: Text document
vm28-975.full
Description: Text document
|