Hello all,
I would like to change the default behavior for our Dell servers
(mostly blades) to stop alerting at all when the watchdog timer
expires. Our HP ProLiant BL460c G1 servers don't alert on timer
expiration. I was hoping to see if there was a difference between the
configs, but the HP servers don't work with ipmi-pef-config ("Unable
to get Number of Alert Policy Entries") and have very few entries in
ipmi-sensors, none of which are related to the watchdog.
What I would like to happen when a watchdog timer expires:
1) The system will reboot
2) *No* SNMP trap sent by the server itself
3) *No* SNMP trap sent by the chassis (if the server is a blade)
4) *No* event inserted in the SEL
5) *No* amber lights on the server or chassis
What I have accomplished:
1) The system will reboot
2) *No* SNMP trap sent by the server itself (the following worked:
"ipmi-pef-config -c -e Event_Filter_17:Enable_Filter=No")
The SEL is populated and an alert sent whether the action is to reboot
the server or do nothing.
What I have tried:
I set everything in "ipmi-sensors-config -S 44_OS_Watch" to be "No":
Section 44_OS_Watch
## Possible values: Yes/No
Enable_All_Event_Messages
No
## Possible values: Yes/No
Enable_Scanning_On_This_Sensor
No
## Possible values: Yes/No
Enable_Assertion_Event_Timer_Expired
No
## Possible values: Yes/No
Enable_Assertion_Event_Hard_Reset
No
## Possible values: Yes/No
Enable_Assertion_Event_Power_Down
No
## Possible values: Yes/No
Enable_Assertion_Event_Power_Cycle
No
## Possible values: Yes/No
Enable_Deassertion_Event_Timer_Expired
No
## Possible values: Yes/No
Enable_Deassertion_Event_Hard_Reset
No
## Possible values: Yes/No
Enable_Deassertion_Event_Power_Down
No
## Possible values: Yes/No
Enable_Deassertion_Event_Power_Cycle
No
EndSection
This changes the output of ipmi-sensors for that host to:
44 | OS Watch | Watchdog 2 | N/A | N/A
| N/A
An unmodified host has this:
44 | OS Watch | Watchdog 2 | N/A | N/A
| 'OK'
After the timer expires, this shows up in the SEL:
ID | Date | Time | Name |
Type | Event Direction | Event
1 | Feb-01-2012 | 07:39:18 | SEL | Event Logging
Disabled | Assertion Event | Log Area Reset/Cleared
2 | Feb-01-2012 | 07:39:23 | OS Watch | Watchdog
2 | Assertion Event | Timer expired, status only
3 | Feb-01-2012 | 07:39:23 | OS Watch | Watchdog
2 | Assertion Event | Timer expired, status only
If I don't disable the SNMP traps from the server for watchdog timer
expiration, I get a trap for DELL-ASF-MIB::asfTrapASRTimeout. A blade
chassis will always send a trap stating that the blade changed from
normal to critical.
Any other ideas? Is this something I need to ask Dell about?
Thanks,
Ryan
--
Ryan Cox
Systems Administrator
Fulton Supercomputing Lab
Brigham Young University
http://tech.ryancox.net