I have the following timeseries entries.
ifDescr{ifDescr="GigabitEthernet1/1",ifIndex="1",instance="x.x.x.x",job="snmp"} 1
ifDescr{ifDescr="GigabitEthernet1/2",ifIndex="2",instance="x.x.x.x",job="snmp"} 1
ifDescr{ifDescr="GigabitEthernet5/3",ifIndex="3",instance="x.x.x.x",job="snmp"}
ifHCInOctets{ifIndex="1",instance="x.x.x.x",job="snmp"}
ifHCInOctets{ifIndex="2",instance="x.x.x.x",job="snmp"}
ifHCInOctets{ifIndex="2",instance="x.x.x.x",job="snmp"}
As is, I have no way of telling which index matches which description, which makes things confusing.
Is there a way to basically to do a join of the above labels using ifIndex to correlate to the the ifDesc label? Or maybe the job can be used to tie the two timeseries together?
I have looked at the group_left feature but haven't been able to figure out how to get it working to combine/aggregate labels.
In this case you want something like rate(ifHCInOctets[5m]) * ignoring(ifDescr) group_left(ifDescr) ifDescr
explanation:
Prometheus will only let you use grouping on operations between series. The value of ifDescr is always "1" so it's safe to multiply.
The ignoring clause means don't use the ifDescr label for matching (as it's only on one of the series). ifIndex, instance and job will be used.
group_left is specifying what labels you want from the series ifDescr. In this case they have the same names.
<vector expr> <bin-op> ignoring(<label list>) group_left(<label list>) <vector expr>