Re: RAID 1

[Matt Howard <mhoward@sigins.com>]

Marshall Lake wrote:
> I'm not sure what to recommend but I know what NOT to recommend.  An
> article concerning the software package named "S.M.A.R.T. Monitoring
> Tools" appeared in Linux Journal a few months back.  (Actually I'm not
> sure if S.M.A.R.T. works with RAID disks.)  This package purported to
> analyze and monitor the health of hard drives.  I ran S.M.A.R.T. on
> the
> hard drive of one of my systems.  Afterwards I had to reformat the
> hard
> drive and reinstall the system in order to get it running again.  I
> contacted the author and it was his opinion that I did something
> wrong. Even if I did do something wrong the software should not
> corrupt my hard drive the way it did.

Yes, SMART is something that can cause problems -- when it was new, many
systems couldn't boot the disks with it on.  For new drives, OTOH, it
can warn you of impending doom.  SMART is very iffy though.

Back to the OP's question:  you can "cat /proc/mdstat" for a quick
overview of raid health.  Also, you can use the "mdadm" tool to monitor
your arrays, and email you if one goes faulty.  (Also, the md driver
will send kernel log messages if an array's state changes.)

I would recommend testing the smart tool on a similar drive.  Do a
"sync" before trying, I have seen old drives that stop when they
receive the SMART commands.  DO NOT USE ANY OF THE SMART TEST
OPTIONS!!! Some disks will read-test, while others will
read/write-test.

A good plan would be to have mdadm monitor your raid, and smartd monitor
your disks.  Are you forwarding syslog to a central logging server? 
Grep the logs for warnings from md.

-- 
Matt Howard <mhoward@sigins.com>
Superior Insurance - Technical Services