RHEL6 or RHEL7 NFS client using kerberos (krb5), one user experiences slow write performance, another does not

Solution Verified - Updated -

Issue

  • NFS is slow for one user on a host that has both sec=sys and sec=krb5 mounts. The kernel only gets into the state after some time with the user using a sec=krb5 NFS mountpoint. Once the kernel is in this state, the mount point with sec=sys is also affected. I don't think that sec=sys mountpoints alone trigger this behaviour.
  • The definition of NFS is slow means NFS mounts have WRITEs capped at PAGE_SIZE bytes/op (4KB on x86_64) even though wsize is much larger - only a reboot fixes the problem.
  • We did not experience this issue prior to upgrading to 2.6.32-504.*el6 kernels

Environment

  • Red Hat Enterprise Linux 6 (NFS client)
    • kernel 2.6.32-489.el6 or above
    • kernel prior to kernel-2.6.32-642.6.1.el6
    • kernel prior to kernel-2.6.32-696.el6
  • Red Hat Enterprise Linux 7 (NFS client)
    • kernel earlier than kernel-3.10.0-514.2.2.el7
    • kernel earlier than kernel-3.10.0-327.46.1.el7
  • NFS mounts with mixed security flavors, some mounts have sec=krb5 and others have sec=sys

Subscriber exclusive content

A Red Hat subscription provides unlimited access to our knowledgebase of over 48,000 articles and solutions.

Current Customers and Partners

Log in for full access

Log In