Sever Life

[tank]

What is an average server's life? Is there such a thing?

Just curious. I'm upgrading server due to some resource things but was wondering how long I could have operated before the HD started giving me issues or something else. I've been on that server for 2 years or so.

[JPC-Les]

Off hand I would say around 2 years is understandable before you see any component failure. However for dedicated servers we replace failed components for you so that’s not really the end of the servers life.

IMHO what will likely bring about the end of your server will be increased resource requirements of new applications and the continuing growth of traffic due to site popularity. Isn’t that what you're seeing now, you had your dedicated two years and now upgrading to a more powerful machine?

I'm sure some of the more venerable members traipsing around the forums who have been working on servers since the 60's may be able to offer interesting perspectives on this.

[Ron]

tank,

Just so that you're aware, a very frequent mode of failure is: after a period of time (years), disk drives frequently fail to spin back up after when you shut down a server for whatever reason. Take backups before shutting down for maintenance!

[JPC-Greg]

There isnt really a set lifespan I would say. Like Les said its really just your needs. Someone running some corporate or low key site on an old p3 could end up sitting there til the cows come home provided replacement parts are handy when something does fail, and it will eventually. The most common scenario is you get a server, in 2-3 yrs if you havent seen a failure and are happy you are probably going to see one very soon. When that failure happens most likely its cheaper and maybe even the only alternative to put uyou back on a modern low end machine instead of hunting down replacements which we probably ebay'd long ago. Its cheaper for us to just stock current parts and keep our stock in rotation. Plus do we really want to trust an old 3yr part we have had shelved and collecting dust.

[tank]

yeah.. those are teh reasons I just opted to upgrade. My main point of concern is the desc it self. Both not spinning back up and failing altogether. Thanks for the info guys. Makes me glad that I'm investing a bit more for peace of mind.

[Pawel Kowalski]

Hard drives have no life span and they are probably the biggest threat you face. I would love to personally run a raid level 5 but can't afford it right now. If you can run it and get 3 hard drives you really don't have to worry about hard drive failures causing you any headaches as if one fails the other 2 still operate and the failed drive can be swapped without any dataloss. The chance of having 2 drives fail at the same time preventing data to be recovered is like 1 in a billion.

If other hardware fails you will suffer some down time but not as catastrophic as if a drive failed.

[JPC-Greg]

Or raid 10 , basically two raid 0's paired up. however I have seen two disc fail in a raid setup before. Its possible , so teh raid 5 1 in a billion is an exageration. Its likely you will be using 3 discs from the same lot, and should one lot be faulty to give you one failure in X months, the others might go to. Its all chance.

In the case where we saw two fail in a raid 10 it was one whole set of jsut one side of the array. Lucky really, since the machine never died and we just put two new and it rebuilt. Had one drive from each side of the array fail it would have all been lost. Which would suck, but is recoverable with backups, etc.

[Ron]

A real danger time is when a machine is brought down for maintenance and the disks spin down and cool; many disks don't like to spin back up after that, so having 2 drives go at once in an array of 4 or 5 drives is probably more like 1 in 100 or 1 in 1,000, but no way is it 1 in a billion.

We've had RAID 5 failures here about 18 months ago that were really painful (to me). Were they the drives or the controllers at fault? I don't know that I was ever told the final determination, but they switched away from the RAID 5.

BOY were they fast; I loaded a million row database, with full individual inserts in under a minute, if I recall correctly. I didn't try the same test once the machine was re-re-rebuilt, may it rest in peace.

[JPC-Greg]

Ron was that in your workplace or you referring to here in jag? We too switched out of raid5 but mostly for performance. Sure raid 10 cost us 25% more since we need 4 drives not 3, and can only make use of 2, but the performance on io and access were just worth it.

[Ron]

I'm talking about here on zeus. I thought I sent you those DB stats when I did it, but I'm sure it was more important to me than to you.

I never got "closure" on what caused all of the early RAID issues, but we started on RAID 5, had several bad crashes, array corruptions and rebuilds of the array. I thought there was a question about the reliability of the RAID 5 Controllers. If you remember from a year and a half ago, it'd be nice if you could end my lack of closure.

As far as spinning drives down, that was at work. I have manually "rotated" disks around their rotational axes and quickly stuffed them back in the drive drawers and powered them up to get them spinning again, before calling for service.

PS That would be a third (33.33%) more for drives since you used to pay $3 and now pay $4. Sorry, it's the programmer in me. Now that you've done it you're stuck with the higher costs; if you go back to 3 drives, you'd only save 25%

I love math humor.