Re: [Ipmitool-devel] supermicro ipmi watchdog issues

[Andy C. <and...@us...>]

Kelsey,

The only software involved is the Linux driver with the ipmi_watchdog
module, and that has been pretty stable for a long time.  IIRC, if one
of the watchdog reset commands returned a bad status, the ipmi driver
should have logged it in /var/log/messages.  

The only option left is a firmware bug in the SuperMicro IPMI firmware.
Support for the IPMI Watchdog timer is Mandatory, not Optional, so they
have to fix it.

Andy

-----Original Message-----
From: Kelsey Cummings [mailto:kg...@co...] 
Sent: Thursday, July 21, 2011 1:09 PM
To: Andy Cress
Cc: ipm...@li...
Subject: Re: [Ipmitool-devel] supermicro ipmi watchdog issues

On Thu, Jul 21, 2011 at 05:59:51AM -0700, Andy Cress wrote:
> The most common problem case is that 1 & 2 are done, but 3 is not, so
> the system resets every time.

;)

> How are you performing these steps?

1 and 2 are being managed via the bios
3 is being managed by the watchdog package which runs on top of
OpenIPMI.

The module options:

IPMI_SI=yes
DEV_IPMI=yes
IPMI_WATCHDOG=yes
IPMI_WATCHDOG_OPTIONS="timeout=120"
IPMI_POWEROFF=no
IPMI_POWERCYCLE=no
IPMI_IMB=no


> It could also be a SuperMicro firmware bug, but it would require more
> detail about the test case to file a firmware bug report.  

They've started to fail after a year or two of being in service.  I'm 
sure this is a hw or software bug and was just looking for some
confirmation.  SM has asked me to fill out a detailed bug report so
their lab can take a look at it now.

-- 
Kelsey Cummings - kg...@co...      sonic.net, inc.
System Architect                          2260 Apollo Way
707.522.1000                              Santa Rosa, CA 95407