Fix a potential error with undefined value in Transport/SNMP.pm

[earendilfr]

Sometimes, I can see in the job queue that some jobs are in a wrong format because we try to call a new method to an undefined object

I suspect that for an unknown reason, sometimes, the $class variable is undefined...

So, I have added a test to validate that the object is defined prior to call the new method

[ollyg]

thanks @earendilfr !
the reason I didn't yet merge this is I'd like to understand why it's happening, as it should not, and maybe the fix is to refactor instead of workaround. thanks for raising the issue, I will take a look.
if you can somehow work out any pattern or reproducing device, do let us know!

[earendilfr]

Hi @ollyg, I will try to reproduce the reason of this error.
One difficulty is the error are not showed in netdisco-backend logs, only in jobqueue through WebUI...

[inphobia]

lemme know what you need, always have some of these in my logs

see this with discover jobs most of the time, in rare cases with arpnip. macsuck i can't see in my current list.

also get this at times, just no message:

one very important note, i'm using:

# discover everything without serials
snmp_field_protection:
  device:
    serial: ['!group:__ANY__']

the ip range tells me this is a 4g device using nbm (network behind mobile. using private apn & provider pushes the routes when it comes online). we have around 2000 such devices & our field team swaps them, often with a different vendor depending on what we have in stock.

[earendilfr]

I confirm I have also tasks in error without any reason when I check:

But I didn't have the config part indicated by @inphobia.

Perhaps an help concerning this incident could be to add, in job queue log, the date and hours where the job was started and finished...

EDIT: forget, I didn't have seen that these informations are available with the "+" icon 😄