Saturday 15 September 2007

SBS - Backup and SS4000-E Intel NAS

Besides using ShadowProtect on our servers, we have a couple of SS4000-E Intel NAS devices to keep live data mirroring on.

Part of our maintenance routine is to use a compressed air source we have in our shop (150psi compressor with dried air) to blow out any equipment we work on.

We had this unit apart to clean out all of the fuzzies and dust. Our office/shop is on the fringes of the city, so there are a lot of fields and bush areas close by. Plus, the construction around us is huge as the area is starting to be developed.

So, we shut down this box and cleaned it out.

When we booted it back up, one of the drives was down. This particular unit has 4 x 300GB Seagates in RAID 5.

We swapped out the "dead" drive and ran the scan on the SS4000. It did not pickup the new drive. A reboot later, and still no sign of the new drive.

We pulled it apart, made sure to give it a good looking over to make sure none of the contacts were dirty and put it back together with the original drive (it tested out okay).

On this boot up, we received a degraded array message, along with a "rebuilding" status for the array. This time around, it had picked up the original drive, but not the data on it.

So, we will keep an eye on this unit to see if it fails the same drive again. If it does, we will swap the unit out for another one with Intel.

The unit is about a year old and has been stable up to this point. We do have a lot of them out there doing similar duty with no issues to date.

There is a firmware update that was released earlier this year that allows for drives larger than 500GB: Intel SS4000-E Downloads Page. Keep in mind that this firmware update will destroy any data on the device.

And, one more point: To rebuild a degraded 900GB RAID 5 array (4x 300GB) on this unit is approximately 17 hours with no other activity to or from the device. Start moving data on or off of the S4000-E and that time quadruples.

Philip Elder
Microsoft Small Business Specialists

*All Mac on SBS posts are posted on our in-house iMac via the Safari Web browser.


stryqx said...

17 hours! That's like just under 5MB/s.
Can you change the rebuild rate, or is that the maximum write performance of the array?
I've never liked RAID-5. Too sucky when it comes to degraded write performance, and that second disk almost always fails with the increased drive workload of a degraded RAID-5.
I prefer RAID-10. Less drive space, but performance is less sucky when degraded and lab testing can be done easily and simply by pulling two drives.
Granted, RAID-5 wins in random read tests, but those writes hurt.

Philip Elder Cluster MVP said...

That is about right. There is a status Web page that one can monitor the rebuild status on. It indicated around 4,700KB/sec when the machine was idle, and about 900KB/sec when I tried to write data to the box.
In most cases, we have a secondary server setup with mirrored data. If we loose an array, we move the users over to that mirror to allow for the rebuild time.
The cost of the extra box and mirroring setup is a lot less than having the whole thing go down.
It is good insurance.