WIP - signal Fleet warning on elastic-agent status

elastic · Sep 8, 2023 · 2a5a605 · 2a5a605
1 parent ac0212d
commit 2a5a605
Show file tree

Hide file tree

Showing 3 changed files with 24 additions and 3 deletions.
diff --git a/internal/pkg/agent/application/coordinator/coordinator.go b/internal/pkg/agent/application/coordinator/coordinator.go
@@ -17,6 +17,7 @@ import (
 
 	"github.com/elastic/elastic-agent-client/v7/pkg/client"
 	"github.com/elastic/elastic-agent-libs/logp"
+	fleetgateway "github.com/elastic/elastic-agent/internal/pkg/agent/application/gateway/fleet"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/info"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/reexec"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/configuration"
@@ -855,8 +856,12 @@ func (c *Coordinator) runLoopIteration(ctx context.Context) {
 
 	case configErr := <-c.managerChans.configManagerError:
 		if c.isManaged {
+			var wErr *fleetgateway.WarningError
 			if configErr == nil {
 				c.setFleetState(agentclient.Healthy, "Connected")
+			} else if errors.As(configErr, &wErr) {
+				// we received a warning from Fleet, set state to degraded and the warning as state string
+				c.setFleetState(agentclient.Degraded, wErr.Warning)
 			} else {
 				c.setFleetState(agentclient.Failed, configErr.Error())
 			}

diff --git a/internal/pkg/agent/application/gateway/fleet/fleet_gateway.go b/internal/pkg/agent/application/gateway/fleet/fleet_gateway.go
@@ -33,6 +33,15 @@ const fleetStateOnline = "online"
 const fleetStateError = "error"
 const fleetStateStarting = "starting"
 
+// Error to be returned when we receive a warning in the Fleet response
+type WarningError struct {
+	Warning string
+}
+
+func (w WarningError) Error() string {
+	return w.Warning
+}
+
 // Default Configuration for the Fleet Gateway.
 var defaultGatewaySettings = &fleetGatewaySettings{
 	Duration: 1 * time.Second,        // time between successful calls
@@ -222,7 +231,12 @@ func (f *fleetGateway) doExecute(ctx context.Context, bo backoff.Backoff) (*flee
 		}
 
 		f.checkinFailCounter = 0
-		f.errCh <- nil
+		if resp.FleetWarning != "" {
+			f.errCh <- &WarningError{Warning: resp.FleetWarning}
+		} else {
+			f.errCh <- nil
+		}
+
 		// Request was successful, return the collected actions.
 		return resp, nil
 	}

diff --git a/internal/pkg/fleetapi/checkin_cmd.go b/internal/pkg/fleetapi/checkin_cmd.go
@@ -77,8 +77,9 @@ func (e *CheckinRequest) Validate() error {
 // CheckinResponse is the response send back from the server which contains all the action that
 // need to be executed or proxy to running processes.
 type CheckinResponse struct {
-	AckToken string  `json:"ack_token"`
-	Actions  Actions `json:"actions"`
+	AckToken     string  `json:"ack_token"`
+	Actions      Actions `json:"actions"`
+	FleetWarning string  `json:"-"`
 }
 
 // Validate validates the response send from the server.
@@ -140,6 +141,7 @@ func (e *CheckinCmd) Execute(ctx context.Context, r *CheckinRequest) (*CheckinRe
 	}
 
 	checkinResponse := &CheckinResponse{}
+	checkinResponse.FleetWarning = resp.Header.Get("Warning")
 	decoder := json.NewDecoder(bytes.NewReader(rs))
 	if err := decoder.Decode(checkinResponse); err != nil {
 		return nil, sendDuration, errors.New(err,