kstore is a simple collector which do list watch with k8s objects and maintain local cache, we can use kstore to track schedule history, and do some statistics, monitoring job on it.
Tips: kstore is still in early and heavily development version, any break or uncompatible changes would be happened.
kstore implemented local cache to maintain k8s object data(currently ONLY pod), and build index for them. we could filter pods by deleted/non-deleted, or POD ip, node name, even time range in the future.
we need to build the kstore bundle, and ensure at least 3 nodes running at one cluster.
$ go build
# single node cluster
$ ./kstore -single -id node-0 -rdir "/tmp/raftdb/node-0" -raddr ":12000" -haddr ":11888" -cluster "127.0.0.1:8080" -config "./kubeconfig" -v 5 -alsologtostderr=true
# multiple node cluster
# node-0
# BUG: leader election still has some issues, node-0 can NOT elect itself as leader unless we set `-single`.
$ ./kstore -single -id node-0 -rdir "/tmp/raftdb/node-0" -raddr ":12000" -haddr ":11888" -cluster "127.0.0.1:8080" -config "./kubeconfig" -v 5 -alsologtostderr=true
# node-1
# join addr should be leader's http address
$ ./kstore -id node-1 -join "127.0.0.1:11888" -rdir "/tmp/raftdb/node-1" -raddr ":12001" -haddr ":11889" -cluster "127.0.0.1:8080" -config "./kubeconfig" -v 5 -alsologtostderr=true
# node-2
$ ./kstore -id node-2 -join "127.0.0.1:11888" -rdir "/tmp/raftdb/node-2" -raddr ":12002" -haddr ":11890" -cluster "127.0.0.1:8080" -config "./kubeconfig" -v 5 -alsologtostderr=true
# with specified index
$ ./kstore -single -id node-0 -rdir "/tmp/raftdb/node-0" -raddr ":12000" -haddr ":11888" -cluster "127.0.0.1:8080" -config "./kubeconfig" -v 5 -alsologtostderr=true -indexes deployName,metadata.labels,DEPLOYMENT_ID
# with sync timeout configured
$ ./kstore -single -id node-0 -rdir "/tmp/raftdb/node-0" -raddr ":12000" -haddr ":11888" -cluster "127.0.0.1:8080" -config "./kubeconfig" -v 5 -alsologtostderr=true -sync.timeout 180