RHEL6: NFS client using automount crashes with 'VFS: Busy inodes after unmount ... Self-destruct in 5 seconds. Have a nice day' prior to the crash

Solution Verified - Updated 2024-08-02T06:50:21+00:00 -

Issue

kernel crashes in clear_inode, iput, or vfs_dq_drop called from shrink_icache_memory
prior to the crash, we see a lot of 'VFS: Busy inodes after umount' messages
kswapd crashes while calling shrink_icache_memory to shrink inode cache
We might increase the automount timeout, as suggested by Why do I see this message in my syslog: "kernel: VFS: Busy inodes after unmount. Self-destruct in 5 seconds. Have a nice day..."?. However, we have around 2000 server with us with exactly same "autofs timeout" value & it doesn't seem to be the problem with any of them.
The issue started after installing 2.6.32-504.3.3.el6.x86_64 version, in our environment all server has 2.6.32-504.3.3.el6.x86_64 and 2.6.18-400.1.1.el5 and all working fine except this server

Red Hat Enterprise Linux 6 (NFS client)
- kernels prior to kernel-2.6.32-573.12.1.el6
- kernels prior to kernel-2.6.32-642.el6
NFSv3 or NFSv4 mounts
NFS server
- an export w/ a sub-dir bind-mounted to the export root
- Originally seen with automount of NFSv3 mount points from datadomain NFS shares

A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.