k8s-event-watcher

This library lets you watch for Kubernetes events, and triggers a callback whenever one matching the criteria is found.

An example of configuration is:

sinceNow: true
filters:
- rules:
    involvedObject.kind: Job
    involvedObject.name: "^*.fail"
    reason: BackoffLimitExceeded
  errorRules:
    # Always mark this event as an error
    type: .*

This would match a Job-failed Event:

apiVersion: v1
kind: Event
...
involvedObject:
  apiVersion: batch/v1
  kind: Job
  name: job-fail
  ...
lastTimestamp: "2019-10-14T07:11:45Z"
message: Job has reached the specified backoff limit
metadata:
  ...
reason: BackoffLimitExceeded
...
type: Warning

Configuration

Configurable keys:

• sinceNow: if true, only processes events generated after the program starts.
• filters: a list of rules. Each rules object is evaluated independently, in an OR fashion. Any event is evaluated on all sets of rules. The first matching filter will cause the trigger of the callback.
• filters.[*].rules: a map of regular expressions. Each regular expression is evaluated against the provided object key. If all the regular expressions match, then the event will be sent to the callback, and the MatchedFields map will be populated with the matching fields.
• filters.[*].errorRules: a map of regular expressions. Each regular expression is evaluated against the provided object key. If all the regular expressions match, then the event will be considered an error, and the MatchedErrorFields map will be populated with the matching error fields.

Local test

You can test the execution of this library by running the example program:

go run ./example --kubeconfig PATH_TO_A_KUBECONFIG_FILE --config ./example/config.yaml

While the example program is running, you can then start a failing job with:

kubectl apply -f ./example/job-fail-k8s.yaml

The example program will then pick up the failure and show the matching event.