Spurious data during DST change hour

[Justin Miller]

Hello,

While preparing statistics for annual reporting on a Lustre filesystem
of ours, I discovered that November 2013 had unbelievably high read
bytes and write bytes in FILESYSTEM_AGGREGATE_MONTH.

The errant data seems to have been logged into the database during the
time that clocks get set back 1 hour for DST. In 2013, that was Nov 3 at
0200.

See attached text file for data gathered by lmtsh from the
FILESYSTEM_AGGREGATE_HOUR table for that time period, plus 1 hour before
and after.

Also attached is a screenshot of lwatch that shows read rate and write
rate had a huge spike then as well.

Do you have any advice on how to drop an hour's worth of data from the
LMT database and recalculate all the aggregate data, or otherwise how
you would handle this?

Versions
Stats server:
LMT Server 3.1.6
Cerebro 1.17
Mysql 5.1.69
CentOS 6.2


MDS/OSS:
LMT Server Agent 3.1.6
Cerebro 1.17
Lustre 2.1.6
CentOS 6.2

Thanks,
Justin Miller