checking for the grid to be 'stable' before starting new droplets

Signed-off-by: Jeff Carr <jcarr@wit.com>

http.go

@@ -85,8 +85,23 @@ func okHandler(w http.ResponseWriter, r *http.Request) {
 
 	if tmp == "/start" {
 		start := r.URL.Query().Get("start")
-		log.Info("should start droplet here", start)
+		log.Info("Handling URL:", tmp, "start droplet", start)
-		fmt.Fprintln(w, "should start droplet here", start)
+		dur := time.Since(me.unstable) // how long has the cluster been stable?
+
+		fmt.Fprintln(w, "should start droplet here", start, shell.FormatDuration(dur))
+		if dur < 17*time.Second {
+			fmt.Fprintln(w, "grid is still to unstable")
+			return
+		}
+		d := findDroplet(start)
+		if d == nil {
+			fmt.Fprintln(w, "can't start unknown droplet", start)
+			return
+		}
+		for _, h := range me.hypers {
+			fmt.Fprintln(w, "could start droplet on", start, "on", h.Hostname, h.Active)
+		}
+
 		return
 	}
 

main.go

@@ -5,6 +5,7 @@ package main
 import (
 	"embed"
 	"os"
+	"time"
 
 	"go.wit.com/dev/alexflint/arg"
 	"go.wit.com/log"
@@ -30,6 +31,9 @@ func main() {
 	readDropletFile("droplets")
 	readHypervisorFile("hypervisor")
 
+	// initialize the grid as unstable
+	me.unstable = time.Now()
+
 	log.Info("create cluser for", argv.Hosts)
 	for _, name := range argv.Hosts {
 		h := addHypervisor(name)

poll.go

@@ -139,15 +139,15 @@ func clusterHealthy() (bool, string) {
 	if me.killcount > 0 {
 		summary += "(killcount=" + fmt.Sprintf("%d", me.killcount) + ")"
 	}
-	last := time.Since(me.stable)
+	last := time.Since(me.unstable)
 	if last > 133*time.Second {
 		// the cluster has not been stable for 10 seconds
-		s := shell.FormatDuration(last)
+		s := strings.TrimSpace(shell.FormatDuration(last))
-		summary += "(unstable " + s + ")"
+		summary += "(stable=" + s + ")"
 	}
 	if good {
-		me.stable = time.Now()
 		return good, "GOOD=true " + summary
 	}
+	me.unstable = time.Now()
 	return good, "GOOD=false " + summary
 }

structs.go

@@ -20,7 +20,7 @@ type virtigoT struct {
 	hypers    []*HyperT
 	droplets  []*DropletT
 	killcount int
-	stable    time.Time // the latest time the cluster has been okay
+	unstable  time.Time // the last time the cluster was incorrect
 }
 
 // the stuff that is needed for a hypervisor