Crash when two resources of the same kind with the same name exist in different namespaces.

[luisdavim]

Hi.

I think I found an issue:

2019/04/15 12:47:10 Checked prometheus endpoint: http://prom-op-prometheus-operato-prometheus.monitoring.svc.cluster.local:9090
2019/04/15 12:47:10 Found ProviderID starting with "azure", using Azure Provider
2019/04/15 12:47:10 Recording prices...
2019/04/15 12:47:10 Interface map[error:found duplicate series for the match group {persistentvolumeclaim="data-vol-kube-opex-analytics-0"} on the left hand-side of the operation: [{persistentvolumeclaim="data-vol-kube-opex-analytics-0", storageclass="standard-disk"}, {persistentvolumeclaim="data-vol-kube-opex-analytics-0", storageclass="<none>"}];many-to-many matching not allowed: matching labels must be unique on one side errorType:execution status:error]. If the interface is nil, prometheus is not running!
panic: interface conversion: interface {} is nil, not map[string]interface {}

goroutine 85 [running]:
github.com/kubecost/cost-model/costmodel.getPVInfoVector(0x1852100, 0xc00058cba0, 0xc0006e2480, 0x0, 0x0)
        /app/costmodel/costmodel.go:642 +0xb20
github.com/kubecost/cost-model/costmodel.ComputeCostData(0x1e088c0, 0xc000424000, 0xc000297680, 0x1e4cca0, 0xc000195110, 0x1b29747, 0x2, 0x0, 0x0, 0x0)
        /app/costmodel/costmodel.go:130 +0xe8b
main.(*Accesses).recordPrices.func1(0xc000424d00)
        /app/main.go:104 +0x476
created by main.(*Accesses).recordPrices
        /app/main.go:101 +0x3f

[dwbrown2]

Good catch, Luis. This is definitely an issue. It actually looks like you were for some reason provisioned a claim that has two storageclasses: standard-disk and <none>. I haven't seen this yet.

Is there any chance you could share the result to this Prometheus query? kube_persistentvolumeclaim_info This should help us pinpoint the issue a little faster.

An easy fix would be to exclude storageclass =<none> but it's not clear this addresses the underlying problem.

[luisdavim]

sorry, I've ended up deleting the extra pvc that had been created by mistake in the default namespce but I can try to reproduce this with some other resource type.
What I had was two pvcs one in the kube-opex-analytics namespace with storage class set to standard-disk and another one with the same name on the default namespace with no storage class set.

[dwbrown2]

Got it, this is helpful. We should now be able to recreate but will follow up if for some reason we can't!

[dwbrown2]

I was able to recreate! PR out for review.