adjacency

Adjacency creates an adjacency matrix for an SRV record to check the latencies between nodes in a network.

It can generate a square (n x n) matrix from all (n) nodes where the service is running. Alternatively, if an SRV record for a different service that runs on m nodes is specified at query-time, it will produce a n x m matrix of latencies from nodes to service endpoints.

Getting Started

Run the adjacency service and specify an SRV record that resolves to all the endpoints where the service is running:

docker run --rm -p 3000:3000 kiloio/adjacency --srv _service._tcp.example.com

Do this for all nodes. Keep in mind, that the value of the --srv flag must to resolve to the endpoints of all the nodes.

Now, to get the square adjacency matrix, make a GET request to the API, e.g. using cURL:

curl example.com:3000 

To get the n x m matrix for the latencies between all nodes of this service and another service's endpoints, specify an srv query parameter in the GET request:

curl example.com:3000?srv=_anotherservice._tcp.example.com

To format the output as JSON, add the format=json query parameter to the request, e.g.:

curl example.com:3000?format=json

Kubernetes

Apply the adjacency service to a Kubernetes cluster with

kubectl apply -f https://raw.githubusercontent.com/kilo-io/adjacency/main/example.yaml

API

/

Generate a complete adjacency matrix:

curl example.com:3000

srv=_service._tcp.example.com

Use the srv query parameter to set the target SRV record to another service. This will result in an n x m matrix.

format

Use the format query parameter to format the output

• format=json JSON
• format=simple only times in a table
• format=fancy table with borders, error code and IP addresses or hostnames (hostname is fallback)
• format=standard error codes with times
• format=svg renders an svg image of the graph

/vector

Generate an adjacency vector from the point of view of a specific node:

curl example.com:3000/vector?srv=_service._tcp.example.com

srv=_service._tcp.example.com

Use the srv query parameter to set the target SRV record to another service.

/ping

Check if service is running:

curl example.com:3000/ping
# The service should respond with pong.