Issue

• We found our server with hundreds or even thousands of stuck netstat and netstat commands and a load average in the thousands, but very little CPU being used
• Processes are getting stuck waiting in the XFS slab shrinker
• We've detected high load on a cluster node and couldn't log in to the system. It was still a member but was unresponsive on the console or over ssh
• corosync is using 100% CPU on only one node, load average is very high, and many processes like ps and netstat seem to be stuck
• While corosync seems to be hogging an entire CPU, there are hung-task warnings in /var/log/messages showing processes stuck waiting in XFS functions
• Why is corosync utilizing so much CPU on one of my nodes?
• We had applications get stuck after processes hung waiting on something, and captured a vmcore. A number of processes are stuck waiting in xfs_fs_free_cached_objects
• We are frequently seeing lvm commands block and LVM resource operations time out.
• corosync and clvmd both spin away with 100% CPU on one node in the cluster