Thursday, March 21, 2019

Streaming Telemetry - What are you monitoring?

I'm in the midst of getting streaming telemetry going in our network and beyond the usual stuff of interface counters, system level processes and BGP RIB route counts I'm beginning to wonder what else people are looking at with ST.

I'd love to hear some ideas on things to monitor!

Background/Setup:

We've been hearing about streaming telemetry from our Cisco SE for some time now but just getting around to looking at it as an augment to our SNMP polling/traps. Sub-5min res and the sexiness of influx/grafana combo is a draw for us. The biggest hurdle was getting the backend systems going and then figuring out the Cisco Pipeline configuration. Our setup is:

  • Docker swarm
  • 3 zookeeper containers
  • 4 kafka containers
  • 1 pipeline container
  • 1 influxdb container
  • 1 grafana container

Cisco XR devices configured to send to Pipeline. Pipeline is configured to send metrics to Kafka and directly to Influx (for now); we'll eventually have influx subscribe to Kafka topics for metrics. If anyone is struggling with the Pipeline setup, I'd be happy to lend some guidance on how I figure out how to translate from the GPB line format to Influx metric format.



No comments:

Post a Comment