'tooBig' error caused by fixed OID limit is going unhandled

[rossmobi]

Bug report

Recently I have setup an InfluxDB server and I am using Telegraf to poll Cisco switches for SNMP traffic. On recent builds of Telegraf, I have seen quite a lot of the following error:

2016/05/17 16:36:32 SNMP Error for host 'XXXXXXX:161': Unable to decode packet: nil

I had also noticed that the series of some switches were simply not receiving metrics, where some switches had received some, sporadically. Most were fine.

I was sure the two were related and after searching in vain, I took a packet capture. I could then see that the switch was returning a 'tooBig' error. My SNMP terminology isn't great, but essentially Telegraf asked for 60 OIDs and the switch returned the error.

The shortest explanation of this error I could find was:

The size of the Response-PDU would be too large to transport.

The value which dictates the maximum number of OIDs in a request appears to be here:

const (
    // maxOids is the maximum number of oids allowed in a Get()
    maxOids = 60

    // Base OID for MIB-2 defined SNMP variables
    baseOid = ".1.3.6.1.2.1"

    // Java SNMP uses 50, snmp-net uses 10
    defaultMaxRepetitions = 50
)

https://github.com/soniah/gosnmp/blob/master/gosnmp.go

Proposal:

I would like to see this constant replaced with a variable in the GoSNMP struct which starts a few lines down from the snippet above.

Use case: [Why is this important (helps with prioritizing requests)]

This is useful because it means legacy or low power/capacity devices which support SNMP can be catered for. It's the client's responsibility to react to and deal with the 'tooBig' error and this way we can.