misc/139039: zpool scrub makes system unbearably slow

[Nathaniel Filardo]

>Number:         139039
>Category:       misc
>Synopsis:       zpool scrub makes system unbearably slow
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon Sep 21 22:30:03 UTC 2009
>Closed-Date:
>Last-Modified:
>Originator:     Nathaniel Filardo
>Release:        9.0-CURRENT
>Organization:
>Environment:
FreeBSD hydra.priv.oc.ietfng.org 9.0-CURRENT FreeBSD 9.0-CURRENT #4: Sun Sep 13 10:12:17 EDT 2009     root at hydra.priv.oc.ietfng.org:/systank/obj/systank/src/sys/NWFKERN  sparc64

>Description:
Even with the sysctl "vfs.zfs.scrub_limit" given value 2, scrubbing my 8-disk array is an exercise in frustration.  The system remains online, and will eventually respond to events, but it's incredibly lagged, even when the events do not directly relate to the controller card and disks making up the array.

The array is 4x320G and 4x750G SATA disks on a
mpt0: <LSILogic SAS/SATA Adapter> port 0x300-0x3ff mem 0x100000-0x103fff,0x130000-0x13ffff at device 1.0 on pci3
mpt0: MPI Version=1.5.10.0

For (I assume unrelated) stability reasons (see pr kern/117688), I have had to run "camcontrol tags $DISK -N 16" for each of the disks in the array.

WITNESS and INVARIANTS are both turned on; I've not tried with them off; should I?
>How-To-Repeat:
Reliably triggered by issuing a scrub request.
>Fix:


>Release-Note:
>Audit-Trail:
>Unformatted: