hard disk drive reliability (fc/sata)
 

Going to have a web site and a couple of servers on a colocated machine (no physical access) and I am a bit concerned about the rebuild time and procedure although we'll use true hardware raid.

From what I read, the weak point is a disk failure during a rebuild of a raid array. Apparently this possibility increases with the capacity of the disks and SATA drives are not as good as fibre channel drives.

Can anyone suggest suitable info? I'd like to be able to find what's most suitable in my case between fibre channel disk drives, SCSI or SATA drives.

The amount of data involved is not very large, 500G drives will do the job on each machine.

I cannot find info concerning FC drives for my sort of set up ( a few drives per machine) and if I have to use SATA drives I wonder if it's possible to have a raid 1 setup and rsync the data to another drive at the same time so that there is 2 copies available in case of drive failure during a rebuild.

There is also raid 6 that looks good.

The servers are going to be mirrored to backup servers but is it practical to restore from a backup server?

Any comment or hint most welcome.

Thank you for your help.

Physically pata and sata drives are USUALLY the same drive with different drive controllers.

The raid limitation will be with your Raid controller. Be aware that there are a lot of fake raid or soft raid controllers out there (most promise controllers are not true hardware raid).

Be aware that support for all FC versions prior to FC5 was dropped about six(?) months ago. Support of FC5 will be dropped when Fedora 7 is released (late April?).

I don't know about practical, but if the system goes down, and restoring from backup is your only option then YES at that point in time, it is very practical. do NOT forgo making backups just because you have build fault tolerance into your system.

My newest systems have been using SAS Drives (Serial Attached SCSI) they are pretty fast and are the newest generation of SCSI ..

This is what I picked up for my Mail serer.. Debian Etch installed without a hitch and detected both the RAID controller and the Network interfaces..
HP DL360 G5 Server and since HP put out a press release that they will support Debian, well thats another reason I chose the HP Server.

also by going this route they have pre-configured servers that will ship same day.. made Dells turn around time look slow in comparison .. Of course you have to order your extra RAM, Hard Drives, low profile DVD-R Drive and rack mounting Rails, all separate.. Still came in under $4,000.00 http://www.provantage.com/hewlett-pa...5~7HEWY0HA.htm

I don't really have too much of a preference when it comes to drives, although I have to say my SAS External 2TB RAID 5 Array kicks the crap out of my 1 TB SATA External RAID 5 Array in performance.. the SAS unit is MUCH faster. This could be caused by a lot of different factors, spindle speed, amount of cache on the drives, command queuing support in the drives, etc.. etc.. The SATA array was much cheaper to build/buy but the SAS Array surpasses it in performance by quite a bit.

Thank you very much for the good info.
I was expecting to make the rsync'ed servers the backup but farslayer seems to suggest that another type of backup is still necessary, I think I must be missing something in that respect, can you elaborate?

Rsync is great, and if you have two servers in different physical geographically separated locations rsync'ing them over the internet might be sufficient. but if both servers are in a single datacenter and it burns to the ground, well then an off site backup would be worth it's weight in gold.

also if both servers you are rsyncing are Internet connected and are exploited, having a known good backup that is offline would be a good thing.

call me overly paranoid if you will, but if you end up in a position with no backup for vital data you will wish you had taken additional steps to ensure you have a clean copy of your data.

I can tell you first hand I have had two servers, One was a database server with our Accounting system database, The second server was the backup, similar to your Rsync scenario.. Both Servers had RAID 5 arrays with Hot spares.

The Backup server had multiple drives fail at the same time and the backup data was lost... OUCH!!! but the production server was still live. I replaced all the failed drives in the backup server array. got it back online but now have to backup ALL the data from the live system. I kicked off the backup job at 5:00 as I left the office. I came into work the following morning to discover the Database server had gone down HARD and also had a multiple drive failure in the RAID 5 Array. Now keep in mind the first server that failed was my backup, and since the DB server Failed During the night it's backup did not complete.. so GAME OVER the database was GONE, and the partial backup was useless..

As luck would have it just before starting the backup job I copied the entire database to my workstation before going home. had I not taken that extra step our company would most likely have gone out of business, or at the very least would have been hurting really bad for a long time. The system was still down for about 2 Days as I had to repair the server, reload the OS, restore the database, and make everything work again..

On the brighter side management FINALLY started to listen to my complaints about our server hardware and backup situation. Since that day last July I have got almost ALL new servers, ALL new workstations, and am working towards a complete overhaul of the datacenter. All things I have been requesting for the last couple years and have been denied time after time. The offsite backup solution was also approved in that budget, as well as many other items I had been asking for. I'm rather annoyed it had to take such a disaster to finally get their attention, but things are MUCH improved now and I can also sleep better at night not worrying about a FrankenServer (read: home built server with Off the shelf parts, that when it goes down you can't find exact replacements for) being in a failed condition in the morning.

I have never before witnessed a RAID 5 Array with 3 drives fail at the same time, let alone TWO separate arrays with that type os failure on two separate servers within a week of each other..

I was literally, physically ill when that DB server crashed, it was a rough couple weeks between the failure of the first server and the recovery of the second and a recovery for me afterwards..

So you can maybe get a little insight to my paranoia.. if it would have only taken $10,000.00 investment to have proper offsite backups, and you forgo that expense and loose all your data and your company goes out of business because of it.. was $10k really that much to spend for insurance ? the Data recovery specialists I spoke with wanted $20k to recover the data from the failed array, and said they couldn't tell if the data was any good until after they recovered it.. "Penny wise, Pound foolish" as they say.

You have to make the determination what your data is worth and if your backup solution is sufficient.

I hope I have given you a little food for thought, that will make your backup solution as robust as it can possibly be, taking into account all possible scenarios.

Best of luck ! :)

Your experience is invaluable, it is the sort of scenario I want to avoid. I had a similar experience in construction a long time ago, but with much less potential for disaster, and I share your paranoia because, I also felt sick at the time. The manager and the senior engineer wanted to insure I got a particular fabrication right and the 3 of us spent far too much time on it just to get it wrong. I felt sick when the goods arrived and checked what had happened.

Thank you again, your story is a reminder that those catastrophes do happen.