NX-OS - EEM Command Failure

Justin Copling · ‎09-29-2015

Hello All,

I have created an EEM script in NX-OS(nexus 6k - 6.0.2) that is returning command failures in the accounting log. Below is the script I am using:

event manager applet python
event syslog pattern "%PFMA-2-FEX_STATUS: Fex .* is online"
action 1.0 event-default
action 2.0 cli python /scripts/fex_ports_disable.py $_syslog_msg
action 3.0 syslog priority informational msg fex ports down

The idea is that this syslog message will execute a python script that is stored in the /scripts/ folder of the bootflash. I am getting the following error in the accounting log:

Sat Jan 19 15:37:30 2013:type=update:id=vsh.9870:user=__eemuser:cmd=python bootflash:/scripts/fex_ports_disable.py (FAILURE)

is there something special that needs to be done in order to allow EEM to pass this variable and run the python script?

Thanks,

Justin

Joe Clarke · ‎09-30-2015

Calling a script from an EEM applet wasn't officially supported until NX-OS 7.x. It has been known to work prior to that, though. The only thing I see that could be problematic is that the syslog message will contain spaces. You might try seeing if a simple Python script can be invoked in this version of NX-OS.

Justin Copling · ‎10-01-2015

So we are able to successfully call the python script from within NX-OS. It just appears to be an issue with getting EEM to wrap the syslog message into the $_syslog_msg variable. When we invoke the python script manually like below, it works successfully.

python fex_ports_disable.py %NOHMS-2-NOHMS_ENV_FEX_ONLINE: FEX-100 On-line

There is logic in the python script to regexp the syslog message for the fex #, and we have tested this.

Do you have any ideas or recommendations on how we can get the syslog message to be set inside the $_syslog_msg variable, or is there another built-in variable that I could send to the python script?

Joe Clarke · ‎10-01-2015

Did you test invoking the script from EEM without arguments? I have no doubt it works manually. I was suggesting you try a simpler test from EEM to see if you can run a Python script at all. As I said, no testing of Tcl/Python execution out of EEM was done until 7.0. I know there have been bugs in the past with this. I do know it's working in 7.x code.

Justin Copling · ‎10-01-2015

Yes, we did..We also changed the EEM script to invoke just the python script like below:

action 2.0 cli python /scripts/fex_ports_disable.py %NOHMS-2-NOHMS_ENV_FEX_ONLINE: FEX-100 On-line

This did in fact work, we could just not get the syslog message to send as a variable correctly.

With your suggestion, I am going to switch the testing to a 5548 running 7.0.6 code and see if we are able to obtain better results.

Thanks!

Joe Clarke · ‎10-01-2015

Hmmm, interesting. Can you paste the output of "show event manager event-types"?

Justin Copling · ‎10-01-2015

Hi Joe,

outputs are below:

usipn-moli19t-sn6klab1# show event manager event-types
Name : oir
Description : OIR event
Event Parameters : devicetype, eventtype, devicenum
Publisher : Platform Manager
Default Action : None

Name : fanabsent
Description : Fan Absent Event
Event Parameters : devicenum, time
Publisher : Platform Manager
Default Action : None

Name : fanbad
Description : Fan Bad Event
Event Parameters : devicenum, time
Publisher : Platform Manager
Default Action : None

Name : memory
Description : Memory Alerts
Event Parameters : memstate
Publisher : Platform Manager
Default Action : None

Name : temperature
Description : Temperature Sensor Event
Event Parameters : modnum, sensornum, major_minor
Publisher : Platform Manager
Default Action : Shutdown on major temperature threshold

Name : allfanabsent
Description : All Fan Absent Event
Event Parameters : devicenum, time
Publisher : Platform Manager
Default Action : None

Name : snmp
Description : SNMP MIB object monitoring
Event Parameters : param1, param2
Publisher : SNMP
Default Action : None

Name : cli
Description : EEM CLI command filter
Event Parameters : command
Publisher : CLI events publisher
Default Action :

Name : link_flap
Description : Link Flaps
Event Parameters : interface, cause
Publisher : Ethernet Port Manager
Default Action : Error Disable the port

Name : storm_control
Description : Traffic on a port exceeds configured thresholds
Event Parameters : interface, cause
Publisher : Ethernet Port Manager
Default Action : None

Name : debug_1
Description : Internal event used by apps to indicate faults.
Event Parameters : interface, cause
Publisher : Ethernet Port Manager
Default Action : None

Name : debug_2
Description : Internal event used by apps to indicate faults.
Event Parameters : interface, cause
Publisher : Ethernet Port Manager
Default Action : None

Name : debug_3
Description : Internal event used by apps to indicate faults.
Event Parameters : interface, cause
Publisher : Ethernet Port Manager
Default Action : None

Name : debug_4
Description : Internal event used by apps to indicate faults.
Event Parameters : interface, cause
Publisher : Ethernet Port Manager
Default Action : None

Name : track
Description : Track object
Event Parameters : object, state
Publisher : Object Tracking Manager
Default Action : none

Name : counter
Description : Named Counter Event
Event Parameters : counter_name, counter_value
Publisher : MVSH
Default Action : None

Joe Clarke · ‎10-01-2015

This is what I suspected. You have the bug where the syslog ED exports the message in the variable $command instead of $_syslog_msg. Change $_syslog_msg to $command and see if that helps.

Justin Copling · ‎10-02-2015

That definitely worked on getting the syslog message into the variable, thanks Joe!

However, I am still getting failures on the execution in the accounting log. Could this partly be from what you spoke of earlier with the python and EEM not being validated yet?

Tue Jan 22 11:47:11 2013:type=update:id=vsh.27607:user=__eemuser:cmd=python bootflash:/fex_ports_disable.py %NOHMS-2-NOHMS_ENV_FEX_ONLINE: FEX-100 On-line (FAILURE)
Tue Jan 22 11:47:11 2013:type=update:id=vsh.27607:user=__eemuser:cmd=syslog priority informational msg fex ports down (SUCCESS)

Joe Clarke · ‎10-02-2015

Yes, it could be. I haven't found any discussion of such a failure internally.