Subscribe
Accepted Solution

Performance Characteristics of Datastores Accessed via FCP and NFS on VI3.5

[ Edited ]

Good Day, 

 

i have seen in the autosupport that there is a bug in our netapp configuration. 

 

Value of the option nfs.ifc.rcv.high is unexpectedly reset

The option nfs.ifc.rcv.high value might unexpectedly get reset to a previous value if it is changed by the option nfs.tcp.recvwindowsize.

 

The option nfs.ifc.rcv.high controls the high-water mark after which NFS level flow control takes effect. This option is also controlled by the option nfs.tcp.recvwindowsize. Changing the option nfs.tcp.recvwindowsize automatically changes the value of the option nfs.ifc.rcv.high.

However, there are some scenarios in which the option nfs.ifc.rcv.high value gets reset to a previous value if it is changed by the option nfs.tcp.recvwindowsize.

To avoid the issue, first configure the option nfs.tcp.recvwindowsize, then manually configure the option nfs.ifc.rcv.high.

Note: The value for the option nfs.ifc.rcv.high should be 1.5 times the value of the option nfs.tcp.recvwindowsize.
 
I am checked my configuartion on the FAS 2220-4 Filer with Ontap 8.2P6.
 
nfs.tcp.recvwindowsize 65940
nfs.ifc.rcv.high 66340
 
 
what setting should at nfs.tcp.recvwindowsize?
 
Greetings 
Stefan
 

Re: Performance Characteristics of Datastores Accessed via FCP and NFS on VI3.5

The default value of 65940 should set to nfs.tcp.recvwindowsize.

Highlighted

Re: Performance Characteristics of Datastores Accessed via FCP and NFS on VI3.5

For this BUG (568415) there is also a note - "The value for the option nfs.ifc.rcv.high should be 1.5 times the value of the option nfs.tcp.recvwindowsize". So if the nfs.tcp.recvwindowsize should be 65940 then nfs.ifc.rcv.high should be 98910.... right?

Re: Performance Characteristics of Datastores Accessed via FCP and NFS on VI3.5

Correct, set:

nfs.ifc.rcv.high=98910

Re: Performance Characteristics of Datastores Accessed via FCP and NFS on VI3.5

Thanks for your help.

Greetings Stefan