spidernet-io / spiderpool

Underlay and RDMA network solution of the Kubernetes, for bare metal, VM and any public cloud
https://spidernet-io.github.io/spiderpool/
Apache License 2.0
523 stars 76 forks source link

add RDMA metrics #3615

Open weizhoublue opened 3 months ago

weizhoublue commented 3 months ago

The Type of Document Issue

Lacking

What's wrong with this document?

projects like nodeExporter could help export metrics of RDMA

node-exporter
https://github.com/prometheus/node_exporter/blob/master/collector/infiniband_linux.go
它的原理是从 /sys/class/infiniband/mlx5_0/ports/1  目录下读取接口

        node_infiniband_state_id{device="mlx5_1",port="1"} 4
        # HELP node_infiniband_symbol_error_total Number of minor link errors detected on one or more physical lanes.
        # TYPE node_infiniband_symbol_error_total counter
        node_infiniband_symbol_error_total{device="mlx5_0",port="1"} 0
        node_infiniband_symbol_error_total{device="mlx5_1",port="1"} 0
        # HELP node_infiniband_unicast_packets_received_total Number of unicast packets received (including errors)
        # TYPE node_infiniband_unicast_packets_received_total counter
        node_infiniband_unicast_packets_received_total{device="mlx5_0",port="1"} 769089
        node_infiniband_unicast_packets_received_total{device="mlx5_1",port="1"} 2.106905e+06
        # HELP node_infiniband_unicast_packets_transmitted_total Number of unicast packets transmitted (including errors)
        # TYPE node_infiniband_unicast_packets_transmitted_total counter
        node_infiniband_unicast_packets_transmitted_total{device="mlx5_0",port="1"} 903967
        node_infiniband_unicast_packets_transmitted_total{device="mlx5_1",port="1"} 3.660728e+06
        # HELP node_infiniband_vl15_dropped_total Number of incoming VL15 packets dropped due to resource limitations.
        # TYPE node_infiniband_vl15_dropped_total counter
        node_infiniband_vl15_dropped_total{device="mlx5_0",port="1"} 0
        node_infiniband_vl15_dropped_total{device="mlx5_1",port="1"} 0
        node_scrape_collector_duration_seconds{collector="infiniband"} 0.081142974
        node_scrape_collector_success{collector="infiniband"} 1
ty-dc commented 2 months ago

macvlan 和 ipvlan