Log EOFs found on Kubernetes watcher at debug level (elastic#10988)

EOFs received when watching Kubernetes events can be recovered by
reconnecting, and they don't need to be harmful. So log them at the
debug level.

CHANGELOG.next.asciidoc

@@ -135,6 +135,7 @@ https://github.com/elastic/beats/compare/v7.0.0-alpha2...master[Check the HEAD d
 - Allow to configure Kafka fetching strategy for the topic metadata. {pull}10682[10682]
 - Using an environment variable for the password when enrolling a beat will now raise an error if the variable doesn't exist. {pull}10936[10936]
 - Add missing `host.containerized` and `host.os.build` to fields.ecs.yml. {pull}11016[11016]
+- Reconnections of Kubernetes watchers are now logged at debug level when they are harmless. {pull}10988[10988]
 
 *Auditbeat*
 

libbeat/common/kubernetes/watcher.go

@@ -255,12 +255,17 @@ func (w *watcher) watch() {
 			r := w.k8sResourceFactory()
 			eventType, err := watcher.Next(r)
 			if err != nil {
-				logp.Err("kubernetes: Watching API error %v", err)
 				watcher.Close()
-				if !(err == io.EOF || err == io.ErrUnexpectedEOF) {
+				switch err {
+				case io.EOF:
+					logp.Debug("kubernetes", "EOF while watching API")
+				case io.ErrUnexpectedEOF:
+					logp.Info("kubernetes: Unexpected EOF while watching API")
+				default:
 					// This is an error event which can be recovered by moving to the latest resource version
-					logp.Info("kubernetes: Ignoring event, moving to most recent resource version")
+					logp.Err("kubernetes: Watching API error %v, ignoring event and moving to most recent resource version", err)
 					w.lastResourceVersion = ""
+
 				}
 				break
 			}