Amazing what companies will throw away. That's how I got this server. I also got a 2nd identical server that I will use for parts if/when stuff fails on this one. Both were runners pulled from sites because of upgrades. Nothing wrong with either one.
I was meaning to ask you about that. I've gotten some corporate hand-me-downs, but nothing approaching something as new and powerful as what you have. The closest I ever came was getting half off on a $10,000 server because I made the purchasing decisions for a $10,000,000 per year client. You are one lucky guy!
Great points all around. The bolded part made me laugh! LOL I'm not that crazy. Just fearful of failures.
The thing about the bolded part is that it's a fairly common practice for large corporations. I wasn't kidding at all. There used to be a company called Comdisco that specialized in colocating mainframe computers at their various facilities around the world for companies like airlines and other businesses that couldn't afford to be down ever. If you have the funds, I can arrange something similar for you as part of my consulting business.
But you made excellent points about RAID5, and since this server does shut off during the day, that will save some drive life. Dunno if I can set this thing up to spin down drives that aren't being accessed.
I wouldn't do that if I were you. I've been a motorsports fan for most of my life, and have learned a thing or two about racing and the cars that are used to do it. Many types of full-on race motors use a device called a pre-oiler to force lubricant oil over bearing surfaces before the motor turns over to reduce friction. This is done because 99% of the wear and tear done on regular car motors takes place in the first few seconds of operation, before the lubricant can circulate to where it's needed. This rule holds true for pretty much anything that spins on bearings, including computer hard drives.
You're not going to find a HD with a pre-oiler, so the worst thing you can do to the longevity of a HD is to spin it up and down often. Premature wear due to lubricant starvation is a well understood cause of failure, but it's not the only one. Heat also plays a role not only in HD life, but in performance as well. Bearing surfaces that are cold can have enough space that they bang against each other until they swell up to their operating tolerances. This can cause galling and premature bearing failure. In addition, as data density grows, modern HDs absolutely must be calibrated from time to time because their tracks will move out of range as minute variances in the size of the platter occur. Because of this most HD storage is kept running constantly in order to reduce wear and maintain thermal stasis.
If you must shut down your machine daily, I suppose you'll need to take extra precautions like doing an incremental backup prior to shutdown. Under no circumstances should you set your drives to spin down if they don't need to. That's strictly an energy-saving move for battery powered devices.
I'm going to re-explore RAID 5.
Please do. RAID5 is the most economical, and IME it's the most you're likely to need as long as you keep a good backup regime. RAID6 adds even more security. Some complain that RAID5/6 is "too slow" compared to massive RAID0 stripe sets. While that may be true, the throughput provided by a properly designed RAID5 array is adequate for all but the most extreme (read: gamer) applications. The cache RAM (that's what the battery's for) on the RAID card helps a lot in boosting real world performance.
Couple more questions: when a drive fails in RAID, do I get an alert through the OS? How does that work?
It depends on how the system is engineered. The cheapest and most common method is to use monitoring software on the computer itself to query the HBA and handle alerts of drive failures, as well as S.M.A.R.T. information that can predict failures before they happen. S.M.A.R.T. is one reason why I am a lot less concerned about having a HD drop dead out of the blue any more.
What happens next depends on what your hardware supports and/or what you want. Back in olden times, Compaq servers had a serial port that could be hooked up to a modem that could dial up a pager or another computer that would then notify someone of the problem. These days there are more sophisticated hardware-based monitoring cards ranging from proprietary designs by the manufacturer to open standards-based
IPMI hardware.
You're stuck with whatever your hardware manufacturer offers (or not). Although IPMI is an open spec., there are no universal IPMI cards that work on anything. Most of these have an Ethernet interface that offer a wide variety of ways to alert the operator, form pager to text message to e-mail to pretty much anything that you can make.
For one system I used an old cellphone taped to the top of the rack, and connected to a serial port on our serial console server. (These were Sun machines.) The console server could send SMS messages through the cellphone to the phones of the technical and management staff, and it could accept incoming calls and connect to any machine. Since the cellphone had its own battery, the whole data center could be blacked out and I'd still have communications.
The hardware method is useful because it can notify you even if the OS has stopped running, which can happen with a severe drive failure. The method that I prefer is the SMS tool that comes with my OpenSUSE distribution. I write my own scripts as glue between my RAID card software and the SMS script.
I'm pretty sure that IBM has its own proprietary system management cards. If yours has one, great. If not, you might balk at the retail price of one, compared to the free price of the machine.
Also, how does one back up 4+TB of data (without a second NAS)?
Tape backup is the most common method in the data center. A tape drive or carousel with the capacity to back up all of your data in a reasonable amount of time isn't cheap, but as you're going to find out, there's a lot more to a complete system than the computer.
Because high capacity tape drives are so costly, and because most home backup products now use an inexpensive albeit slow external HD, I supplement my own cross-backup product with inexpensive external drives or arrays. I've been using a 2TB USB drive to store a duplicate of the 2TB data partition on my Windows workstation that runs the software that turns my TiVo recordings into unencrypted MPEG files. I use an inexpensive ($1K) Netgear NAS appliance to store a duplicate of the contents of my file server and my Linux workstation, and an IEEE-1394B connected external HD to backup my Linux workstation.
I'm not going to go into the details about my own cross-backup product because that's how I put food on the table.
I will say that it makes the semi-portable USB and IEEE-1394 redundant to the point that when I'm on the road I carry one or more of them for convenient access to my data without being bound to a slow Internet connection.
