inventory operator has too long of GPU labels

[Zblocker64]

When using a rtx 4080 super, the label is too long for k3s/k8s to handle

ERROR	nodes.node.monitor	couldn't apply patches for node "uzinode"	{"error": "Node \"uzinode\" is invalid: metadata.labels: Invalid value: \"akash.network/capabilities.gpu.vendor.nvidia.model.rtx4080super.interface.pcie\": name part must be no more than 63 characters"}

Assumption is we need to rework how the labeling is done.

Maybe akash.network/gpu.nvidia.rtx4080super.interface.pcie

Other possibility is to name super cards rtx4080s s for super

However this would still cause issues for a card like the rtx4000sffada

[uzi-akash]

Thank You 🥇 Akash Team
This was moved to: #233

[cloud-j-luna]

When using a rtx 4080 super, the label is too long for k3s/k8s to handle

ERROR	nodes.node.monitor	couldn't apply patches for node "uzinode"	{"error": "Node \"uzinode\" is invalid: metadata.labels: Invalid value: \"akash.network/capabilities.gpu.vendor.nvidia.model.rtx4080super.interface.pcie\": name part must be no more than 63 characters"}

Assumption is we need to rework how the labeling is done.

Maybe akash.network/gpu.nvidia.rtx4080super.interface.pcie

Other possibility is to name super cards rtx4080s s for super

However this would still cause issues for a card like the rtx4000sffada

I wouldn't close this issue.
GPU codes could be part of the value in the label, but then it would require a list of GPU codes for the "GPU capability", there is a limit of 63 characters in the value as well which I believe was the reason for the current labelling.
Is there any historical context available as to why the choice was to use labels instead of annotations? Annotations are ideal to store metadata which seems to be the case in these specific "capability labels".

[uzi-akash]

This was moved to: #233