Custom probe/sensor classifies but doesn't identify

[scott111]

Writing custom probe for HP Enclosure device. My custom probe allows unsupported shells and does "show enclosure info" ssh command and gets back plenty of info to classify. Discovery output using custom probe says it classifies based on ssh output but then it goes ahead and tries SNMP to classify and fails. I've read that any sensor failure marks the whole discovery as a failure. Discovery output:

SNMP Classify                   SNMP probe timed out. Target is either unreachable or there are no valid credentials for it.

SNMP Classify                   Active, couldn't classify, now finished

UNIX Classify                   Unsupported shell, '$0', and probe parameter 'allow_unsupported_shells' is set to 'false'. Supported shells are [ksh, sh, bash, tcsh, csh]

UNIX Classify                   Active, couldn't classify, now finished

UNIX Classify                   Active, couldn't classify, now finished

UNIX Classify                   Classified, now Identifying

Shazzam                                   Active, now Classifying

ShazzamLaunch         MID midhost.com discovering 10.xx.xx.xx

Discovery                             Discovery started

Debug "gs.log" statements print from my classifier and from custom sensor but not from my identifier.

[ismailkiswani]

Check the ECC Queue (Input and Output XML)

[ismailkiswani]

Check the ECC queue entries and check if SNMP returned the sysObjectID.

if returned check it in the SNMP OID table

for HP enclosure you should create   custom probe, and than create a parameter called allow_unsupported_shells and set that value to true. This resolved the issue, and I was to able to test the probe and run some commands.

[ismailkiswani]

Hi Scott,

This might help you same issue

HP BladeSystem c7000 Enclosures

[scott111]

The Bladesystem has a non-standard cli and so yes you need to allow unsupported shells and I had figured that out early on.

The SNMP probe returns:   <results probe_time="12995"><result error="SNMP probe timed out. Target is either unreachable or there are no valid credentials for it."

We may re-visit discovery through SNMP at some point but right now can I tell the discovery to NOT use SNMP? Apparently, if any one method (snmp, ssh, wmi, etc.) fails then the whole discovery of the target is marked failed.

"... there is code in the sensor that interprets any error as a failure and it triggers additional classifier even though the class matched". From the link I posted above.

[scott111 · Accepted solution]

Problem solved. I simply needed to add a Bladesystem - Identity probe that was triggered by my Bladesystem classifier. Thanks for the info!