On Feb 9, 2011, at 2:51 AM, Matt Connolly wrote: > Thanks Richard - interesting... > > The c8 controller is the motherboard SATA controller on an Intel D510 > motherboard. > > I've read over the man page for iostat again, and I don't see anything in > there that makes a distinction between the controller and the device. > > If it is the controller, would it make sense that the problem affects only > one drive and not the other? It still smells of a drive issue to me.
It might be a drive issue. It might be a controller issue. It might be a cable issue. The reason for the slowness is not necessarily visible to the OS, beyond the two queues shown in iostat. > Since the controller is on the motherboard and difficult to replace, I'll > replace the drive shortly and see how it goes. The hardware might be fine. At this point, given the data you've shared, it is not possible to identify the root cause. We can only show where the slowdown is and you can look more closely at the suspect components. Lately, I've spent a lot of time with LSIutil and I am really impressed with the ability to identify hardware issues on all data paths. Is there a similar utility for Intel controllers? > Nonetheless, I still find it odd that the whole io system effectively hangs > up when one drive's queue fills up. Since the purpose of a mirror is to > continue operating in the case of one drive's failure, I find it frustrating > that the system slows right down so much because one drive's i/o queue is > full. Slow != failed, for some definition of slow. -- richard _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
