2

Running on LSi 9260-16i, with 16 SAS drives (Seagate Savvio 15k.2). One drive is getting media read errors, so we're going to RMA it out. Which of these is the least performance destroying option in the meantime until the new drive gets here in a day or two?

  1. Leave the drive in, ignore the media errors, and just swap it out in a week when the new drive arrives. This isn't really an option, but it's worth listing to cover all possibilities. Boss and I agree we're not going to do this.

  2. Take drive offline in management software, and let array function as a degraded set. I've never actually done this, so I'm not sure of the performance hit we might take. Understandably, this way is sort of risky.

  3. We have a very similar drive from another (unused) server, that is similar to the other 15 drives, but is 3Gbps instead of 6Gbps. This is where my question comes in; will this performance hit be MORE or LESS than taking a drive offline?

All the 16 drives work together as a RAID10 array with one big volume. I'm leaning towards leaving the 3Gbps in and taking the hit, but I was wondering what other people think would be best here.

jski
  • 911
  • 1
  • 7
  • 20

1 Answers1

2

Let the disk fail or run until you can replace it. The controller is smart enough to complete reads from its pair in the interim.

You can also use a 3Gbps disk in a pinch without consequence. Rotating/spinning disks don't come anywhere near to saturating there 6Gbps or 3Gbps links, so there's no issue there. You're likely running those 16 disks on an oversubscribed expander backplane, so the disk link speed won't make a difference in this case.

ewwhite
  • 197,159
  • 92
  • 443
  • 809
  • Wow. We've been using very quick SSD drives for most of our servers lately (this is one of the real old ones), and I didn't even think about the fact that spinners wouldn't come close to the throughput limits. Thanks for the assist! – jski Oct 01 '14 at 14:16