Ditch RAID and just go to HBAs with a better file system like ZFS or CEPH.
----- Mike Hammett Intelligent Computing Solutions Midwest Internet Exchange The Brothers WISP ----- Original Message ----- From: "Dev" <[email protected]> To: [email protected] Sent: Tuesday, December 5, 2017 11:48:06 PM Subject: Re: [AFMUG] RAID controller eats disks? Old RAID controllers are notorious for misbehaving and giving false readings on drive/volume health. Additionally, the older the controller, the harder it is to find good shelf spares with matching model/firmware, etc. that will be happy reading your RAID volume(s) in the event of a controller failure. This is why I’ve been slowly migrating to a reasonably tolerant - like RAID 6 - array which is configured with software RAID in the native OS (typically Linux in my case). Nothing wrong with hardware RAID, it just can be a little tricky to recover data in the event the hardware dies. With the speed of hardware these days and size of drives, in a 4-drive array using RAID 6 you can build some very large volumes with great fault tolerance on vanilla hardware with good I/O performance, and even if the whole server turns into a smoking hole, you can rebuild your data on some other standard hardware. Also, as drives increase in capacity, it’s more important to have more than just one volume for parity, two is nice in RAID 6. You might be able to build a new box and start migrating your data in case you have a failure and difficulty getting your data. Hardware is usually cheaper than the data on it, and peace of mind is nice. Good luck whichever way you choose, hope you keep all the important data. >> > I've got a somewhat old Dell Poweredge with a PERC H700 RAID controller. About a year ago SMART predicted a failure on disk 4, so I replaced it. A few weeks ago SMART predicted a failure on disk 4, so I replaced it. Today SMART predicts a failure on disk 4. On the second incident I have no doubts, because the disk made audible noises. I'm just curious why it's always disk 4. Can the controller conceivably do something that harms the disk? Just a statistical anomaly? It's a RAID 1+0 by the way, so there should be a nearly identical workload on one of the other disks.
