Power Supplies are Turned off after boot

Mike Jones proclivis at gmail.com
Sat Aug 8 09:00:43 AEST 2020


I managed to get rid of the watchdog service by removing:

IMAGE_FEATURES_remove = “obmc-net-ipmi obmc-host-ipmi” where the latter had a dependency on phosphor-watchdog.

But the supplies still shut off. journalctl has no entries with the keyword “watchdog”. grepping for power shows the chassis-power* services.

I confirmed that each device receives an OPERATION command via PMBus, so clearly something knows the devices are there via devicetree/sysfs.

What are all the services capable of sending PMBus commands related to power?

Mike

> On Aug 6, 2020, at 8:32 PM, Milton Miller II <miltonm at us.ibm.com <mailto:miltonm at us.ibm.com>> wrote:
> 
> On August 5, 2020 Mike Jones wrote:
> 
>> Now that I have webui displaying telemetry for multiple PMBus
>> devices, I want to solve a power supply problem.
>> 
>> After boot, all PMBus devices have their power turned off via an
>> OPERATION command. This happens about 10-20 seconds after boot
>> completes. There is a message printed saying it is disabling them,
>> and one line per action. Like:
>> 
>> VOUT 1: disabling
>> 
>> Where VOUT 1 does not match the name of the rails in the config files
>> with the name and limits. It seems like a more generic term
>> indicating voltage.
>> 
>> These devices are in the device tree and have hwmon running.
>> 
>> What service would be capable of shutting off the supplies? It would
>> have to be able to discover them, perhaps by device tree, then issue
>> standard PMBus commands.
> 
> I'm guessing the host watchdog was not shutdown when your HOST 
> completed boot and the watchdog requesed the system power off.
> Normally an IPMI message from the host will disable the watchdog.
> 
> You should be able to see the transition files run in the systemd 
> journal if this is the case.
> 
> if not try watching for activity by using journalctl to follow 
> events live in a shell.
> 
>>> Note: the Phosphor System Manager service fails at start up. I don’t
>> know if this matters, I was going to figure out what it does later.
>> 
>> Mike

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ozlabs.org/pipermail/openbmc/attachments/20200807/4208f6ea/attachment-0001.htm>


More information about the openbmc mailing list