RHEL6.6: NFS client not caching just written to NetApp Clustered NFS server (C-DOT) when file is larger than a certain size

Issue

• We notice that when we read a just written NFS file back from the same client it is not read from the client cache.
• This problem is seems very similar to a previous problem we experienced which is documented in RHEL 6.5 NFS client does not cache just written files on an NFS share mounted from a NetApp filer, but there are the following two differences:
  • We use cDOT (clustered Data Ontap) 8.3 instead of 7-mode filers
  • Small files are cached but larger files are not.
• By 'small' I mean - on the system I tested I found that up to 46K, files are read from NFS client cache but above that files are read from the filer.
• I did not test if there is difference in upper limit, so above a certain amount the files are always read from the filer.
• In the traces you see:
  • Create a directory and write a file (46k or 47k)
  • I wait around 5 second and then I read the file for the first time.
  • You can see the NFS V3 read call
  • I wait around 5 seconds again and read the file for the second time.
  • I wait around 5 seconds and the file is deleted again.
• We run the kernel (2.6.32-504.8.1.el6.x86_64) that is supposed to have the fix for RHEL 6.5 NFS client does not cache just written files on an NFS share mounted from a NetApp filer