Re: Software RAID (was [OT] 19"/2U Cases)
Michael Loftis wrote:
> --On August 30, 2007 6:31:40 PM +0800 Thomas Goirand <firstname.lastname@example.org>
>> As well, for monitoring, I suggest you to monitor the disks with smartd
>> (apt-get install smartmontools). You will need an MTA to receive the
>> warnings. This helps a lot when having a lot of server to remove disks
>> before they crash the system.
> Have you seen SMART successfully predict or even notice an issue in
> production? We have drives go out all the time, many/most with SMART
> monitoring and smartd enabled, but never have we had SMART notice
> anything going wrong. And I'm talking more than just single read
> errors, drives that became complete bricks without warning from SMART.
> I don't have much faith in it as I've not yet seen it work and predict
> any failure at all.
With the hardware I use (ICH6 from Intel on supermicro boards), smartd
and mdadm both yell in my mailbox when a HDD has DriveSeek errors.
Of course, SMART is not magic, it is said that it "predicts" failure,
but it's not god and wont guess when your drive will fail.
However it doesn't always send a mail before the system crashes, the
fact that sometimes you wont notice that a system has one of it's drive
marked faulty makes it a very important report tool. Otherwise you never
notice when a drive is broken, and that's how you get 2 drives broken at
the same time...
Also, note that smartd can monitor the health of your hard drive behind
a 3ware hardware raid if you like. My friend Damien did it on one of our