This document is intended to provide contributors with an introduction to develop GM(GlobalManager) of the Sedna project.

Debug GM

  1. config GM:

The config file is the yaml format:

kubeConfig: ""
namespace: ""
websocket:
  port: 9000
localController:
  server: http://localhost:9100

  1. kubeConfig: kubernetes config file, default ""

  2. namespace: the namespace GM watches, "" means that gm watches all namespaces, default "".

  3. websocket: since the current limit of kubeedge(1.5), GM needs to build the websocket channel for communicating between GM and LCs.

  4. localController:

    • server: to be injected into the worker to connect LC.

Generate a config yaml:

cat > gm.yaml <<EOF
kubeConfig: "${KUBECONFIG:-$HOME/.kube/config}"
namespace: ""
websocket:
  port: 9000
localController:
  server: http://localhost:9100
EOF

  1. compile and run GM directly:

If you have already run Sedna by following the install doc, you need to stop GM by kubectl -n sedna scale --replicas=0 gm before, and reconfig GM_ADDRESS of LC by kubectl -n sedna edit daemonset lc.

make WHAT=gm
_output/bin/sedna-gm --config gm.yaml -v4

Alternatively you can debug GM with golang delve:

dlv debug cmd/sedna-gm/sedna-gm.go -- --config gm.yaml -v4