Slashdot Mirror

← Back to Stories (view on slashdot.org)

Fibre Channel and RAID Arrays

Posted by Cliff on from the tracking-the-elusive-problem dept.

reptyle asks: "Anyone have useful experience with Fibre Channel RAID arrays? I deal with some Seagate Barracudas and one or another fails once in a while, but with no regularity. Before you ask: Yes, they are RAID5 compliant. The problem here is that I have trouble tracing the origin of the failure; it's not always the same hard drive; it's not necessarily the encasement or the adapter, and it's not clear what it is. But I suspect the particulars are less interesting than general collective testimonials of experience. I read that FC-AL is the kick-ass hot commodity, but I'm having problems seeing the performance boosts over the SCSI that were promised. I've read the FAQS, installed the latest firmware and drivers, and replaced individual hard drives, all to no avail. Any insights?"

8 comments

  1. Slow RAID 5 fibre? Check mode pages by draziw · · Score: 1

    Check the drive mode pages. Things that might need to be changed: Write Cache Enable (WCE) should be 1, Read Cache Disable (RCD) should be 0, read/write retries (even at zero, drives have hidden retries, and some drives just then use a default number) try 0 if you want for speed, but drives will fall out when they are really good, and number of cache segments (many drives default to 3, you might do better with 2.

    Ryan

    1. Re:Slow RAID 5 fibre? Check mode pages by Anonymous Coward · · Score: 0

      >Write Cache Enable (WCE) should be 1 Probably not a good idea if you don't have a UPS or some form of cache battery backup. Unless you like losing data if there's a power failure, that is.

    2. Re:Slow RAID 5 fibre? Check mode pages by draziw · · Score: 1

      If they are running fibre drives in a RAID 5 config, I figure they aren't a moron, and are running on a UPS.

    3. Re:Slow RAID 5 fibre? Check mode pages by Anonymous Coward · · Score: 0

      Never underestimate the power of ignorance.

    4. Re:Slow RAID 5 fibre? Check mode pages by Anonymous Coward · · Score: 0

      First of all, is this by chance a Sun A5x00 RAID box? If so, you need to hit Sunsolve, and check on ALL of the relevant patches and firmware upgrades for the disks, FCAL interface boards in the array, and in the machine. Voice of experience here, a mismatch of firmware rev in ONE disk in the array can bring an entire loop down.

      As far as hardware failures on the drives, I've seen quite a few failures involving the loop interface circuitry on the drives. Basically there's a set of op amps that amplify the signal and pass it to the next drive in the array along the backplane. And since the setup is basically a token ring, if one fails, the whole loops dies a horrible flaming death.

      In short, make sure your firmware and patches are up to snuff.
      In terms of performance, it seems that FCAL disk systems tend to perform best with large data transfers across the bus, such as database reads/writes, while SCSI tends to be better for small block transfers. Someone out there with a bit more knowledge care to expand on why?

      Also, anyone out there had any experience getting an A5000 working on an UltraSparc with a PCI FCAL card in it, running Linux? Been meaning to try this in the lab at work, but haven't had the time yet...

    5. Re:Slow RAID 5 fibre? Check mode pages by draziw · · Score: 1

      Ah - One more thing I forgot to mention. If it's ok to have the wait vs a drive failing; set your SCSI card, and the driver to allow at least 8 seconds for (recovered errors/Drive Command Timeouts). - That's how long these guys take to recover.

  2. Re:Last! by Anonymous Coward · · Score: 0

    So nerr, nerr, ne nerr nerrrrr! :p

  3. Same problem here by Lev_Arris · · Score: 1

    >But I suspect the particulars are less interesting than general collective testimonials of experience.

    Don't say that. We've got exactly the same problem here and we're still trying to figure out what's going on. At first we assumed that it was due to the enclosures. Then we suspected the cause to be power failures (because we found out that one side of the mirror was not connected to the UPS systems) I just read your posting and I find this very interesting because we happen to have Seagate Barracudas here too.

    We'll keep investigating it.