Symptom
As soon as switch is removed from DCNM, it becomes slow to respond and the sysinfo service crashes. Contrary to the log message, a core file may not be saved.
%SYSMGR-2-SERVICE_CRASHED: Service "sysinfo" (PID 29474) hasn't caught signal 6 (core will be saved).
Switch# run bash
bash-4.3$ ps aux | grep nginx
(Thousands of nginx_f worker sessions are printed in the format below)
svc-nxa+ 5892 0.0 0.0 296772 5692 pts/14 Ss+ 15:21 0:00 nginx_f worker
bash-4.3$ copy /volatile/nginx.log /bootflash/
bash-4.3$ exit
Switch# show file bootflash:/nginx.log
(Thousands of these entries are generated)
_pterm_create_vsh_session:291 pid:5879 User sa-dcnm does not exists!
pterm_get_vsh:810 pid:5879 couldn't create a vsh session
Conditions
Switch was removed from DCNM, however the DCNM tracker feature was not disabled before doing so.
Workaround
Disable the DCNM tracker feature in DCNM, before deprovisioning the switch.
If the switch is already in this state, destroy the guestshell and reload the switch:
Switch# guestshell destroy
Switch# reload
Further Problem Description
Problem is not observed if the switch is downgraded to 9.3(3). Reinstalling the guestshell does not reintroduce the issue.
