Thanks for the tip. In the local case, I could send to the iSCSI-backed ZFS RAIDZ at even faster rates, with a total elapsed time of 50seconds (17 seconds better than UFS). However, I didn't even both finishing the NFS client test, since it was taking a few seconds between multiple 27K files. So, it didn't help NFS at all. I'm wondering if there is something on the NFS end that needs changing, no? Also, how would one easily script the mdb command below to make permanent?
On 5/5/06, Eric Schrock <eric.schrock at sun.com> wrote: > My gut feeling is that somehow the DKIOCFLUSHWRITECACHE ioctls (which > translate to the SCSI flush write cace requests) are throwing iSCSI for > a loop. We've exposed a number of bugs in our drivers because ZFS is > the first filesystem to actually care to issue this request. > > To turn this off, you can try: > > # mdb -kw > > ::walk spa | ::print spa_t spa_root_vdev | ::vdev -r > ADDR STATE AUX DESCRIPTION > ffffffff82dc16c0 HEALTHY - root > ffffffff82dc0640 HEALTHY - /dev/dsk/c0d0s0 > > ffffffff82dc0640::print -a vdev_t vdev_nowritecache > ffffffff82dc0af8 vdev_nowritecache = 0 (B_FALSE) > > ffffffff82dc0af8/W1 > 0xffffffff82dc0af8: 0 = 0x1 > > > > See if that makes a difference. > > - Eric > > -- > Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock >
