Installing RAID Controller

[MikeH30]

I have a Dell PowerEdge SC 1430 (Dual Core Intel® Xeon® 5110; 4MB Cache, 1.6GHz, 1066MHZ FSB) server with a 160GB hard drive. This is my production box. I would like to add another 160GB hard drive with RAID 1. I am looking to have a mirrored back up of hard disk 1.

It is my understanding that installing the hard drive and RAID 1 controller card is pretty straight forward. I am currently running SLES10-SP2 and plan on upgrading (via a clean install ) to SLES11 in the next few weeks.

1. Is it better for me to wait and install the additional hard drive and RAID 1 controller during a fresh install of SLES11?

2. Do I really need to purchase my 2nd hard drive and RAID 1 controller from Dell ( PERC 5/I SAS internal RAID adapter, PCI-Express ) ? Since this is my production box I thought that it might be better to get my parts from Dell in order to minimize compatibility issues. RAID is new for me.

Thanks,

Mike

[gogalthorp]

You definitely want to do RAID at the install.

Hardware RAID can be a problem. If in a year the RAID controller fails, you can have problems if you try to replace with a different model.

You might want to think about software RAID.

In any case RAID is not really a backup option.

[MikeH30]

Gogalthorp,

It sounds like hardware RAID is not dependable.

My goal is to minimize down time in the event of a system failure. Our main business applications are web based. I currently back up the web root to a USB external hard drive via Cron. Our user base is 10 or less with no more than 5 users on at any given time. If we had a system crash today, I would have to reinstall the OS and then restore the web root files. Rather than focus my energy on RAID, should I look to utilizing Zmanda and remote back up?

[HROAdmin26]

Hardware RAID is very dependable - why else do you typically find real HW RAID cards in high-end servers? Hardware RAID cards fail *much less often* than HDD's themselves. *If* a HW RAID card fails though, your *guaranteed* recovery option is to attach the HDD's to another card of the same model (or a compatible model as specified by the manufacturer's testing.)

That's why non-CORP users many times opt for software RAID via the md driver - recovery options are a little more flexible.

RAID is not a backup - it is an HA solution. A failed HDD does not bring down your server. It will not protect against file deletion, bad users, bad/corrupt data, etc. - that is what backup does.

If you want a hybrid data backup/minimum downtime solution, you can use a tool like rsync to mirror the active drive to the other *at timed intervals.* So if you rsync once a day, that's your backup - 1 copy, once a day. Recovery time is improved though in a DR scenario because you can move the backup drive > primary quickly.

My goal is to minimize down time in the event of a system failure.

That is what RAID is for - keeping the volume [notice I said volume and NOT HDD] online (HA). It WILL NOT give you any "backup." A backup plan should be implemented separate from any RAID concerns. RAID makes it *less likely* that you will need to perform a *DR* restore using the backups due to a failed HDD. That is all.

[gogalthorp]

I did not mean to imply hardware RAID is undependable. The problem is that RAID controls tend to have propritary bits. This means that in event of a controller failure you may need to replace with the exact same controller model and manufacturer. If you have good backups this is not really a problem.

[MikeH30]

Is it advised to buy two RAID controllers at the same time to avoid issues with matching manufacturer/model in event of a failure? With how rapidly models change, the odds of getting the same one a year later could be slim.

[gazj]

For a small system like yours I would opt for software raid rather than over complicating things. Software works really well and really doesn't take much cpu cycle. As has been said before raid is not a backup solution, merely a means to keep your system running during hard disk failure.

Look into getting a scsi tape drive (DDS4) as an on site backup solution you can alternate tapes for the day of week giving you seven days of data to fall back on. Rsync is great for mirroring data offsite. (I tend to run this every hour)

The best piece of advice I can give regarding backups, is to prove a restore works about once a month. There is nothing worse than trying to restore and finding your backups have not been working correctly (even if software has said they have). I am talking from an unfortunate exp here.

Good Luck

[MikeH30]

Gogalthorp, HROAdmin26, Gazj:

Thanks for all of your expert input on this topic. I realize now that I need a combination of RAID ( to keep my system running during a crash) and a solid back up solution. I am going to look into software RAID. I found a nice tutorial at:

Installing SLES on Software RAID1 | Novell User Communities

My ideal back up plan would be to mirror data offiste. Gazj mentions utilizing Rsync for this.

Thanks for pointing me in the right direction guys!!

-Mike

[gazj]

The bash script I call once an hour in cron (this is on the machine that contains the mirror using ssh to login to the main machine)

cron line = 30 * * * * /usr/bin/mirror-swj >> /var/log/mirror-swj

This keeps records of the rsync in /var/log/mirror-swj (adjust file names and script etc to suit your needs and lookup the rsync manual for the options suitable for your mirror)

#!/bin/bash
date
if ! [ -f /var/run/mirror-swj ]
        then
        touch /var/run/mirror-swj
        rsync -rtlvh -e ssh --delete swj@1xx.1xx.2x.1x:/swj /srv/mirrors
        else
        echo "The last mirror update is still running exiting"
        exit
        fi

rm /var/run/mirror-swj
date
echo
echo -----------------------------------------
echo

[gazj]

Oh the tutorial looks fine by the way, never used sles but the instructions look similar to the cli way of setting up software raid in a gui sorta way, lol, so I'm sure you will be fine