Improve SNMP plugin error logging

[kostasb]

It would help with troubleshooting if errors mentioned particular failing OID's in the SNMP plugin.

E.g. appending the OID that failed to the error log would help identify problematic ones: https://github.com/influxdata/telegraf/blob/master/plugins/inputs/snmp/snmp.go#L448

[phemmer]

Agree. Its a trivial change. I'll try to get to it tonight, and see if there's any other errors which can benefit from more context.

[Ebrink]

It would also be helpfull if the errorlog would contain the specific SNMP instance and/or device that is failing. Every once in a while I get a "snmp collection took longer than collection interval", but it is hard to find out which instance or device produces the error with 53 SNMP instances polling 400+ devices.

[kostasb]

@Ebrink The took longer than collection interval message is generated by the Telegraf agent itself, which is an external component to the snmp plugin: https://github.com/influxdata/telegraf/blob/master/agent/agent.go#L174 . It is thrown when the agent times out a plugin's execution because the interval is reached.

I will check how more verbose output can be provided there.

[kostasb]

@phemmer For the OID, I think it would suffice to append the string to the message as "performing get on %s",oid. It would also be useful to specify the actual agent that encountered the polling error.

[kostasb]

@phemmer @sparrc We are still logging some messages, particularly full-on timeouts without an agent / OID to identify what the failure was about:

2016/12/13 12:31:15 E! ERROR: input [inputs.snmp] took longer to collect than collection interval (5m0s)

[phemmer]

I have a local branch where I've done most of the work on this, just need to finish it up (mostly just make sure I didn't miss anything). I'll see if I can get to it this weekend.

[phemmer]

Sorry for the huge delay on this. For as trivial as a change as it was, I should have finished it up long ago :-(

Anyway, PR is up (#2220). However the PR does not address the most recent comment (the "took longer to collect than ..."). This error comes from telegraf core, and not the snmp plugin. Trying to address this could be tricky as inputs don't really have a standard "name" identifier. The only way I can think to address it is to add the line number of the config where the input begins.

[kostasb]

Thanks @phemmer for the PR.

As for the took longer to collect than ... error messages, I believe @sparrc plans on adding a naming scheme for plugins: #1815