CONTAP-777838: ONTAP node disruption after BSD wired memory grows with fix for CONTAP-520139
Issue
An ONTAP node may experience a disruption after a userspace process such as mgwd, secd, or bcomd remains unresponsive and cannot recover:PANIC: Process <process_name> unresponsive for <seconds> seconds ... in process nodewatchdog
A generic node watchdog signature is not sufficient to match this issue. The disruption event must also have the FreeBSD memory-exhaustion signature:
- FreeBSD wired memory consumes nearly all of the BSD memory partition.
All known field hits show approximately 94% or more wired memory. - Available BSD memory is reduced to a few hundred MB or less.
- Systemshell
top -bSoutput shows highWiredand littlefreememory usage.
Systemshell access requires diagnostic privilege. Use
systemshell commnd only with NetApp Support guidance to determine current MHOST memory usage using the top command, example:cluster::> set -privilege diagnostic
cluster::*> systemshell -node <node_name> -command "top -bS | head -n 5"
last pid: 78952; load averages: 29.70, 28.44, 26.02 up 138+18:44:26
3339 processes:72 running, 282 sleeping, 2 zombie, 2983 waiting
CPU: 2.3% user, 0.0% nice, 36.9% system, 0.1% interrupt, 60.7% idle
Mem: 62M Active, 1132K Inact, 3532K Laundry, 15G Wired, 905M Buf, 253M Free
Swap: 8192M Total, 1400M Used, 6792M Free, 17% Inuse
