RHEL6.3: kernel hang / soft lockup under heavy NFS4 load, events kernel thread stuck in umount_tree holding vfsmount_lock
Issue
- While testing the nfsv4 client provided in RHEL 6, the machine displays a load of 2000 in top for a while and suddenly the server reboots without any warning.
- Same test script runs fine on a different NFS client machine
- The /var/log/messages files are getting filled with "soft lockup" errors.
- High nfs thread count / nfsd thread count
Environment
- Red Hat Enterprise Linux 6.2 - 6.4
- Seen on RHEL6.2 / 6.3 kernels prior to kernel-2.6.32-358.18.1.el6
- Not specific to NFS version
- Thousands of NFS automounts
- Seen on RHEL NFS client to Solaris OpenIndiana NFS server over NFSv4
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.