monitoring: start a new workstream on common metrics across vendors

[glimchb]

• Marvel @prasunkapoor taking action item to propose common metrics from Marvel side they feel comfortable
• Nvidia @AnuradhaKaruppiah taking same action item
• Intel @dandaly will help find a relevant contact for us

[mestery]

Worth mentioning ipdk-io/openconfig-public#1, as it is a PR on virtual device statistics exposed via openconfig.

[glimchb]

thanks @mestery for that comment, that PR is completely un-aligned with that OPI tries to produce

[dandaly]

Hi Boris can you elaborate?

[dandaly]

For monitoring what we are supporting are two sets of schemas for stats in IPDK:

1. openconfig-interfaces: There are a set of statistics that can be found here: https://github.com/ipdk-io/openconfig-public/tree/master/release/models/interfaces
2. virtual-devices: These are virtual devices (networking and storage) that use the same set of statistics as openconfig-interfaces. These devices are different from openconfig interfaces but from a statistics point of view we have made extra effort to align them. They are documented here: https://github.com/ipdk-io/openconfig-public/blob/master/release/models/virtual-devices/openconfig-virtual-devices.yang

This is meant to work for both software targets like KVM (combination of kernel, vhost-user and vfio-user dataplanes) and hardware targets like DPUs, IPUs & switches.

[AnuradhaKaruppiah]

Here are the statistics published via the DOCA Telemetry service -
https://docs.nvidia.com/doca/sdk/doca-telemetry-service/index.html#description

In addition the following ethtool counters are supported -
https://support.mellanox.com/s/article/understanding-mlx5-ethtool-counters

Boris, you talked about starting with system level counters? I am not sure if that is covered by these two groups. Can you please give me an example of that and I can dig up that info.

[glimchb]

@AnuradhaKaruppiah this is a good starting list with system level metrics here #5