strange disk load distribution

[Svlad]

Hi,

I am rather puzzled by the load distribution (attached below). All disks below are in a single raidz2 unit, and all except one show identical response. da8, however reports a significantly lower write volume. My puzzlement is a) how this is possible, and b) whould I no longer trust that disk, the reporting stats, or both?

I wonder if anyone here has encountered similar riddles before?

 

[Ericloewe]

what's the output of zpool status and smartctl -a /dev/adaX (for all drives)?

Use pastebin, please.

 

[DrKK]

Yes, please, let's see the zpool status and the smartctl's.

I am curious to resolve this one myself.

 

[Svlad]

Thanks for the replies. I followed your suggestion and suspect I have hound the issue: da8 is not a raidz2 member, but rather an old 3ware array that for some reason was added as a stripe to the zpool. It's quite odd; I remember deselecting it from the list of devices since I wanted to break it into JBOD and set up another zpool on those disks. But at any rate this solves the current mystery. It's still remarkable how abysmal that hardware raid performance is.

Thanks for your help!

 

[cyberjock]

I don't mean to single you out Svlad, but there is one thing I find hilarious about the hardware RAID crowd (and i'm not saying you are in this group Svlad).

So the regular readers and posters see people every day or two that argue that hardware RAID is just the worst with ZFS. If you tell them there's a 50% penalty their head would explode and they'd ditch hardware RAID in a heartbeat. But if you tell the it's not really compatible they'll argue until the end of time. Maybe we should just tell them all that you'll kill performance of the pool by 50-80%. No joke, that's what I saw in my tests when I was first learning about FreeNAS.