SNMP plugin needs a new schema

[yellowpattern]

• Version of collectd: 5.5.3
• Operating system / distribution: CentOS 6

Sometimes it is possible to collect interface statistics using SNMP, in other instances it is done with collectd. Unfortunately the two methods do not present statistics in an aligned name space. To fix this, the SNMP plugin needs to be updated.

In the SNMP configuration, we can setup collecting octet counters like this:

<Data "Octets">
    Table true
    Type "if_octets"
    #InstancePrefix "octets"
    Instance "MIBNAME::portName"
    Values "MIBNAME::ifInOctets" "MIBNAME::ifOutOctets"
</Data>

What actually happens is that the SNMP plugin applies the "Instance" name to the "type_instance" field rather than the "instance" field.

What I'd like to suggest is the addition of a new per-Data boolean option "NewSchema" that when set to true, causes collectd to use "instance" rather than "type_instance", will cause "InstancePrefix" to be ignored and add a new configuration option, "TypeInstance" that allows the "type_instance" field to be set.

The goal here is to make it easier to re-use queries and graphs between interface and SNMP statistics. The "type_instance" change is to introduce another 'variable' to use with grafana when doing "GROUP BY".

[faxm0dem]

we should probably add support for metadata to snmp

[rpv-tomsk]

We can allow to override reported plugin value on per-Data basis,
and can add an option PluginInstance in addition to existing Instance.

All this allow to unify metrics from interface and snmp plugins.

[rpv-tomsk]

This feature was implemented. If you still interested, you can do tests.
Please inform us if you will do this, we will wait for feedback then.