Hi
We are seeing a strange performance problem on some of our NFSv3 clients:
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
1.6 0.0 0.6 0.0 37.6 8.8 23506.7 5526.7 100 100
filer:/vol/mbox4004
0.2 0.0 0.0 0.0 0.0 3.2 0.0 16137.7 0 100
filer:/vol/mbox4001
3.0 0.2 3.6 0.0 3.7 8.7 1159.9 2718.5 100 100 filer:/vol/mbox4002
Looking at the mountpoint with fsstat shows:
new name name attr attr lookup rddir read read write write
file remov chng get set ops ops ops bytes ops bytes
46.5M 9.50M 12.8M 6.46G 149K 32.3G 72.3M 91.0M 2.74T 50.9M 246G /mbox4001
2 0 0 1.64K 0 8.19K 17 2 34 1 17 /mbox4001
3 0 0 2.92K 0 14.6K 22 1 18 1 17 /mbox4001
1 0 1 2.90K 0 14.5K 18 23 14.5K 8 60.3K /mbox4001
What we see is a high number of "attr get" and "lookup ops". Almost no data is
read or written.
I know there is a high cpu load on the NetApp filer, and we are looking into
that problem as well.
I've read somewhere, that NetApp's ONTAP getattr handling is a single-threaded
process. So, my clients might be hammering the filers to death :-)
Is there any way to lower the lookups a little bit? I know about the mount
options, but I don't want to turn a knob without knowing why...
Thanks for your help!
--
This message posted from opensolaris.org
