Solved: Is QoS compatible with pNFS in cDOT8.3 and cDOT8.3.1?

Dong_Chen · ‎2016-01-18

IHAC is test QoS in cDOT831P1. I found the TR-4211 of performance primer cDOT8.3(Page18)

"QoSpolicies are applicable to all supported protocols, including NFS, SMB, SAN, iSCSI, and FCoE. Starting in clustered Data ONTAP 8.3, NFS 4.1 is supported.

Note: QoS is not compatible with NFS 4 prior to clustered Data ONTAP 8.3.

QoS is not compatible with pNFS in clustered Data ONTAP."

So I have a test of QoS and pNFS. It seems that the QoS policy group works in the pNFS access. Both in the direct LIF mount or the indirect LIF mount, the QoS throughput did limit the IO.

So, what does the incompatibility between QoS and pNFS ly in, the limitation or something else? Or QoS is supported with nNFS?

Here are the output,

step 1, enable nfsv4,v4.1,v4.1-pnfs,

cluster1::*> nfs show -fields v4.0,v4.1,v4.1-pnfs -vserver svm1

vserver v4.0 v4.1 v4.1-pnfs

------- ------- ------- ---------

svm1 enabled enabled enabled

cluster1::*> vol show -fields qos-policy-group -volume vol1

vserver volume qos-policy-group

------- ------ ----------------

svm1 vol1 qp1 -> QoS policy group “qp1” with the limitation of 1MB/s

cluster1::*> qos policy-group show

Name Vserver Class Wklds Throughput

---------------- ----------- ------------ ----- ------------

qp1 svm1 user-defined 1 0-1MB/s

[root@rhel1 /]# mount -t nfs -o vers=4.1 svm1:/vol1 /vol1 -> mount with 4.1 option in RHEL6.5 host

[root@rhel1 /]# mount

svm1:/vol1 on /vol1 type nfs (rw,vers=4,minorversion=1,addr=192.168.0.140,clientaddr=192.168.0.61)

[root@rhel1 /]#

[root@rhel1 vol1]# dd if=/dev/zero of=q3 bs=1M count=30

30+0 records in

30+0 records out

31457280 bytes (31 MB) copied, 70.9384 s, 443 kB/s -> dd test the file creation while the rate without QoS limitation is 145 MB/s

[root@rhel1 vol1]# dd if=/dev/zero of=q2 bs=1M count=300

30+0 records in

30+0 records out

314572800 bytes (315MB) copied, 272.228 s, 1.2 MB/s

cluster1::*> qos statistics volume performance show -> qos statistics display the throughput is around 1MB/s

Workload ID IOPS Throughput Latency

--------------- ------ -------- ---------------- ----------

-total- - 0 0KB/s 0ms

-total- - 15 981.33KB/s 1.69s

vol1-wid8129 8129 15 981.33KB/s 1.69s

-total- - 17 1109.33KB/s 3.91s

vol1-wid8129 8129 17 1109.33KB/s 3.91s

-total- - 6 362.67KB/s 2.56s

vol1-wid8129 8129 6 362.67KB/s 2.56s

-total- - 0 0KB/s 0ms

-total- - 13 810.67KB/s 374.49ms

vol1-wid8129 8129 13 810.67KB/s 374.49ms

-total- - 17 1088.00KB/s 1.91s

vol1-wid8129 8129 17 1088.00KB/s 1.91s

Dong_Chen · ‎2016-01-27

Thanks for the replying, I have got the confirmation that NFS 4.1 and pNFS are now supported in clustered Data ONTAP 8.3.1+. In clustered Data ONTAP 8.3 only NFS 4.1 was supported.

View solution in original post

Renifa · ‎2016-01-27

Hi,

pNFS support began in clustered Data ONTAP 8.1 for files only and continues with enhancements in clustered Data ONTAP 8.2. There is no Data ONTAP 7G/7-Mode support for pNFS. Current client support
for pNFS is very limited, but NetApp does support all clients that support pNFS and follow the RFC specifications. By default the pNFS option is enabled, but it is only active if NFSv4.0 and NFSv4.1 support also are enabled. NetApp does not currently recommend pNFS for metadata-heavy workloads. pNFS requires a client that also supports pNFS. Currently, RHEL 6.4 is the only commercial Linux distribution that has full pNFS support. Check the link for more info "http://www.cptech.com/wp-content/uploads/2014/06/Clustered-Data-on-Tap.pdf"

Thanks

If this post resolved your issue, help others by selecting ACCEPT AS SOLUTION or adding a KUDO.

Dong_Chen · ‎2016-01-27

Thanks for the replying, I have got the confirmation that NFS 4.1 and pNFS are now supported in clustered Data ONTAP 8.3.1+. In clustered Data ONTAP 8.3 only NFS 4.1 was supported.