topic Re: How to detect a Raid failure with Linux ? in Operating System - Linux

How to detect a Raid failure with Linux ?

Mike_811 — Thu, 24 Jul 2008 12:24:31 GMT

Hi guys,
I've a DL380 running Linux Centos. The OS is installed on disks mirrored with a controller RAID0+1.
How can I setup the OS in order that when the mirror is broken an alarm raises (snmp, Email, logs...) ?
I can't see anything with the ILO and in the logs ?
Thanks for your help

Re: How to detect a Raid failure with Linux ?

Heironimus — Thu, 24 Jul 2008 15:06:40 GMT

On a Proliant you should install the HP tools for the Red Hat Enterprise Linux version that corresponds to your installed CentOS version. You'll probably need to edit /etc/redhat-release to look like an official RHEL version to get the tools installed and configured, but once they're set up you can put the CentOS one back and they should run just fine.

I think the HP agents send SNMP traps, but you will need some other tool to handle those traps and do something useful.

Re: How to detect a Raid failure with Linux ?

Ivan Ferreira — Thu, 24 Jul 2008 20:25:58 GMT

The HP insight management agents can send you a mail when something goes wrong with the hardware:

This is a mail I received recently:

Trap-ID=3034

Logical Drive Status Change: Slot 0, Drive: 1.
Status is now Rebuilding.

Re: How to detect a Raid failure with Linux ?

dirk dierickx — Tue, 29 Jul 2008 05:28:05 GMT

i always put my money on the software raid from linux itself.

Re: How to detect a Raid failure with Linux ?

skt_skt — Tue, 29 Jul 2008 16:33:47 GMT

i heard of Raid Aplication CDs which can be used to install raid management s/w. But not sure if they are available for Linux versions.

Re: How to detect a Raid failure with Linux ?

Ralph Grothe — Fri, 01 Aug 2008 09:48:03 GMT

Sorry, if this doesn't apply to your HW RAID layout.
But I also would second Dirk in his preference for Linux's own SW RAID.
We have been running all our Linux servers with this for several years now and never have experienced any data loss owe to some RAID failure.
I think the Linux MD layer is ultra stable,
very easy to set up and administer.
The same goes for the mdadm monitor mode
which lets you most easily plug in your own custom alerting or event handling scripts (I for instance have it send passive check results to my Nagios server, which always has notified me in time when there needed a disk to be replaced).
On RHEL you are already provided with an mdmonitor init script.
In /etc/mdadm.conf all what's left is to set PROGRAM to point to your custom event handler.
e.g.
# grep ^PROGRAM /etc/mdadm.conf
PROGRAM /usr/local/sbin/mdevent.pl