Chassis/Host power control and power state transition

Chassis/Host power control and power state transition

[rgrs]

[-- Attachment #1: Type: text/plain, Size: 578 bytes --]

Hi All,

Sometimes when I do "ipmitool chassis power cycle",  system gets powered off but not powered on back again.
Logs show some sd_bus_set_property returns errno 5 (Input/Output error).

My guess is that CPLD that takes care of power distribution  in my server is causing the IO error, but i need enough data/logs to prove it.

I'm not able to debug further without understanding the state machine, system targets, and their order of execution.
Is there any docs to tools to help me understand or visualize the execution sequence (for power operations atleast)?

Thanks,
Raj

[-- Attachment #2: Type: text/html, Size: 937 bytes --]

Re: Chassis/Host power control and power state transition

[Andrew Geissler]

> On Dec 20, 2019, at 3:03 AM, rgrs <rgrs@protonmail.com> wrote:
> 
> Hi All,
> 
> Sometimes when I do "ipmitool chassis power cycle",  system gets powered off but not powered on back again.
> Logs show some sd_bus_set_property returns errno 5 (Input/Output error).
> 
> My guess is that CPLD that takes care of power distribution  in my server is causing the IO error, but i need enough data/logs to prove it.
> 
> I'm not able to debug further without understanding the state machine, system targets, and their order of execution.
> Is there any docs to tools to help me understand or visualize the execution sequence (for power operations atleast)?

https://github.com/openbmc/phosphor-state-manager/blob/master/README.md and
https://github.com/openbmc/docs/blob/master/architecture/openbmc-systemd.md are
good places to start.

Unfortunately we have not had a lot of luck with visualization. I have an issue,
https://github.com/ibm-openbmc/dev/issues/944, to look into doing something
with systemd-analyze but haven’t made much progress.

Usually when debugging failed systemd services, systemctl and journalctl
are going to be your debug tools.

Andrew

> 
> 
> Thanks,
> Raj
>