I'm thinking of using unRAID in a production environment for cloud storage and I before I started doing some testing I was wondering how reliable is unRAID?

 

I'm going to be setting it up with 3 2TB drives to start and I need something that is going to be working 24/7.

 

Thanks

I hardly ever turn my server off. Works well in most situations.

 

I'm building one for work to host monthly archives.

I hardly ever turn my server off. Works well in most situations.

 

I'm building one for work to host monthly archives.

My older unRAID server has been in operation for 5 1/2 years, 24/7, and except for an intermittent power splitter and a drive that failed, it has worked perfectly.

Most of the software is Slackware Linux.

For backups, archives, etc. from a low number of users it is very reliable.  For an enterprise environment in which many users need to be able to access it at once, it isn't the best choice.

For backups, archives, etc. from a low number of users it is very reliable.  For an enterprise environment in which many users need to be able to access it at once, it isn't the best choice.

 

For me only 1 server will be accessing it to start. I'm using it as an expanding cloud storage setup for a large image/file hosting site that pushes anywhere from 20 - 120 Mbps.

In short, only as reliable in the amount of time you spend maintaining it, and the reliability of each hard drive, and server component.

The ability for unRAID to recover (and not destroy data) is the strong point.

 

 

http://lime-technology.com/forum/index.php?topic=1751.0

 

Unraid as a software is very reliable. However a reliable software should also run on very reliable hardware and IMHO in the last few months or a year the attention was turned to the "eye candy" side of the hardware - you know how important is to have hot-swappable 5x3 and all that nice cabling jobs.

 

The real important things - ECC memory, UPS protection are not that glamorous or even mentioned at all. There is a reason why the industry use them and there is a reason why we have here "unexplained" errors. I will quote here from the AMD white papers that the expected incidence of the bit(s)flips is one bitflip per 2-4 weeks per

gigabyte of DRAM. To put this another way, a 4GB system can expect to encounter roughly one error each week.

 

This does not mean that we have to shell big bucks like the big corporations. The plain vanilla unbuffered ECC memory is only marginally more expensive (10% or so) that the regular ones. The AMD functionality is build in the CPU itself and you just have to pick a motherboard that supports it. The Intel camp is normally way more expensive as Intel wants to make big bucks from these "features" but you can use Intel only if you need the IPMI functionality.

 

Add a cheap APC UPS and you will have a very reliable system.