Help: Host powered ON and after seconds Host Power OFF script is started.

Help: Host powered ON and after seconds Host Power OFF script is started.

[A, Maheswari]

[-- Attachment #1: Type: text/plain, Size: 2359 bytes --]

Hi,

I had basic understanding of below processes and its dependency in openbmc.

button_power.exe
power_control.exe ( without timeout and interval for GPIO Pgood pin polling ).
Chassis_control.py
System_manager.py ( which is again depends on skeleton/configs/Platform.py.
                                       In my case I have created one dummy Platform.py by referring existing openbmc platform).

Issues is observed in below two scenarios:

  1.  Whenever I am pressing power button.
  2.  Whenever sending REST command to Power On the Host. (RequestedHostTransition --> ON).

Issue:

    In both scenarios I am able to power on the host, But We are getting watchdog Timeout and,
   after some time target is reached quiesce state, and Host power OFF script is also started Running.
   (Linked host-poweroff script in  obmc-host-stop.target.requires ).

    Also phosphor-state-manager is always getting Host state as HOST_STATE_QUIESCE_TGT in sysStateChange() functionality.



  1.  What are all the process is responsible to send Host state signal as quiesce to phosphor-host-state-manager. ?.
  2.  If phosphor-watchdog or system_manager is responsible.Please let us know clear view of understanding.
  3.  In our platform, We are not using BOOTCOUNT sensor concept as initial phase.

Please refer short journal log:
Nov 09 05:38:04 HOSTNAME phosphor-watchdog[1338]: Timer Expired
Nov 09 05:38:04 HOSTNAME systemd-journald[640]: Forwarding to syslog missed 11 messages.
Nov 09 05:38:04 HOSTNAME phosphor-watchdog[1338]: watchdog: Timed out
Nov 09 05:38:04 HOSTNAMEi systemd[1]: Created slice system-op\x2docc\x2ddisable.slice.
Nov 09 05:38:04 HOSTNAME systemd[1]: Detaching egress BPF program from cgroup failed: Invalid argument
Nov 09 05:38:04 HOSTNAME systemd[1]: Starting OpenPOWER OCC Active Disable...
Nov 09 05:38:04 HOSTNAME systemd[1]: Reached target Quiesce Target.
Nov 09 05:38:04 HOSTNAME systemd[1]: Stopping Phosphor poweron watchdog...
Nov 09 05:38:04 HOSTNAME systemd[1]: Stopped Phosphor poweron watchdog.
Nov 09 05:38:04 HOSTNAME phosphor-host-state-manager[1215]: Auto reboot enabled, rebooting
Nov 09 05:38:04 HOSTNAME phosphor-host-state-manager[1215]: Beginning reboot...
Nov 09 05:38:04 HOSTNAME phosphor-host-state-manager[1215]: Host State transaction request

Thanks
Maheswari A

[-- Attachment #2: Type: text/html, Size: 12303 bytes --]

Re: Help: Host powered ON and after seconds Host Power OFF script is started.

[Lei YU]

> What are all the process is responsible to send Host state signal as quiesce to phosphor-host-state-manager. ?.
> If phosphor-watchdog or system_manager is responsible.Please let us know clear view of understanding.
> In our platform, We are not using BOOTCOUNT sensor concept as initial phase.
>

You are hitting the issue because of watchdog timeout.

Behind that there are several pieces:
1. phosphor-watchdog@poweron.service is started when host is powered on;
2. During host boots, hostboot sends ipmi watchdog reset commands to BMC;
3. If the watchdog times out, the service will start
obmc-host-timeout@0.target, and the "Wanted" services will be started,
including obmc-host-quiesce target.