Re: [Smartmontools-support]Using Smartmontools in a cluster

[Bruce A. <ba...@gr...>]

Hi Jeffrey,

>    I've been looking at using smartmontools on a few clusters
> especially since I read Bruce's comments that he uses it in his
> clusters. I've been trying smartmontools on a few different hard
> drives (some old, some new) to get a feel for what is being reported.
> Now, I'm ready to start diving into parsing smartmontools output for
> useful information to watch for drive problems and drive failures.

What we do on our cluster is:

(1) Run smartd, with this config file:
# First and second ATA/IDE hard disk.  Monitor all attributes
/dev/hda -S on -o on -a -I 194 -m me...@my...dress
/dev/hdc -S on -o on -a -I 194 -m me...@my...dress

(2) Run self-tests once per week from a cron script:

#! /bin/bash

# Once per week, run extended self-tests on the disks see man smartctl
# for further details of how this works.

if [ -e /proc/ide/hda ] ; then
    /usr/sbin/smartctl -t long /dev/hda > /dev/null 2> /dev/null && \
	/usr/bin/logger -t rundiskselftests "Starting long self-test on
/dev/hda" || \
	/usr/bin/logger -t rundiskselftests "FAILED starting long
self-test on /dev/hda"  
fi

if [ -e /proc/ide/hdc ] ; then
    /usr/sbin/smartctl -t long /dev/hdc > /dev/null 2> /dev/null && \
	/usr/bin/logger -t rundiskselftests "Starting long self-test on
/dev/hdc" || \
	/usr/bin/logger -t rundiskselftests "FAILED starting long
self-test on /dev/hdc"
fi

(3) We have a separate script running that periodically uses smartctl -A
and grep/awk to monitor the temperature of the disk, and sends an email if
they are too high.

> However, let me ask, does anyone have any scripts for parsing the
> output of smartmontools for a cluster setting? 

I'd suggest using smartd for this.

On another cluster that I've heard about, they run smartd, and then run a
SHORT self-test about once every hour or so.

Cheers,
	Bruce