Add trace support for decommissioning (#15502)

* Add trace support for decommissioning
* Add support for tracing errors during decommission

cmd/decommetric_string.go

@@ -0,0 +1,25 @@
+// Code generated by "stringer -type=decomMetric -trimprefix=decomMetric erasure-server-pool-decom.go"; DO NOT EDIT.
+
+package cmd
+
+import "strconv"
+
+func _() {
+	// An "invalid array index" compiler error signifies that the constant values have changed.
+	// Re-run the stringer command to generate them again.
+	var x [1]struct{}
+	_ = x[decomMetricDecommissionBucket-0]
+	_ = x[decomMetricDecommissionObject-1]
+	_ = x[decomMetricDecommissionRemoveObject-2]
+}
+
+const _decomMetric_name = "DecommissionBucketDecommissionObjectDecommissionRemoveObject"
+
+var _decomMetric_index = [...]uint8{0, 18, 36, 60}
+
+func (i decomMetric) String() string {
+	if i >= decomMetric(len(_decomMetric_index)-1) {
+		return "decomMetric(" + strconv.FormatInt(int64(i), 10) + ")"
+	}
+	return _decomMetric_name[_decomMetric_index[i]:_decomMetric_index[i+1]]
+}

cmd/erasure-server-pool-decom.go

@@ -800,11 +800,14 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 						logger.LogIf(ctx, err)
 						continue
 					}
+					stopFn := globalDecommissionMetrics.log(decomMetricDecommissionObject, idx, bi.Name, version.Name, version.VersionID)
 					if err = z.decommissionObject(ctx, bi.Name, gr); err != nil {
+						stopFn(err)
 						failure = true
 						logger.LogIf(ctx, err)
 						continue
 					}
+					stopFn(nil)
 					failure = false
 					break
 				}
@@ -819,6 +822,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 
 			// if all versions were decommissioned, then we can delete the object versions.
 			if decommissionedCount == len(fivs.Versions) {
+				stopFn := globalDecommissionMetrics.log(decomMetricDecommissionRemoveObject, idx, bi.Name, entry.name)
 				_, err := set.DeleteObject(ctx,
 					bi.Name,
 					encodeDirObject(entry.name),
@@ -826,6 +830,7 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 						DeletePrefix: true, // use prefix delete to delete all versions at once.
 					},
 				)
+				stopFn(err)
 				auditLogDecom(ctx, "DecomDeleteObject", bi.Name, entry.name, "", err)
 				if err != nil {
 					logger.LogIf(ctx, err)
@@ -881,6 +886,47 @@ func (z *erasureServerPools) decommissionPool(ctx context.Context, idx int, pool
 	return nil
 }
 
+//msgp:ignore decomMetrics
+type decomMetrics struct{}
+
+var globalDecommissionMetrics decomMetrics
+
+//msgp:ignore decomMetric
+//go:generate stringer -type=decomMetric -trimprefix=decomMetric $GOFILE
+type decomMetric uint8
+
+const (
+	decomMetricDecommissionBucket decomMetric = iota
+	decomMetricDecommissionObject
+	decomMetricDecommissionRemoveObject
+)
+
+func decomTrace(d decomMetric, poolIdx int, startTime time.Time, duration time.Duration, path string, err error) madmin.TraceInfo {
+	var errStr string
+	if err != nil {
+		errStr = err.Error()
+	}
+	return madmin.TraceInfo{
+		TraceType: madmin.TraceDecommission,
+		Time:      startTime,
+		NodeName:  globalLocalNodeName,
+		FuncName:  fmt.Sprintf("decommission.%s (pool-id=%d)", d.String(), poolIdx),
+		Duration:  duration,
+		Path:      path,
+		Error:     errStr,
+	}
+}
+
+func (m *decomMetrics) log(d decomMetric, poolIdx int, paths ...string) func(err error) {
+	startTime := time.Now()
+	return func(err error) {
+		duration := time.Since(startTime)
+		if globalTrace.NumSubscribers(madmin.TraceDecommission) > 0 {
+			globalTrace.Publish(decomTrace(d, poolIdx, startTime, duration, strings.Join(paths, " "), err))
+		}
+	}
+}
+
 func (z *erasureServerPools) decommissionInBackground(ctx context.Context, idx int) error {
 	pool := z.serverPools[idx]
 	for _, bucket := range z.poolMeta.PendingBuckets(idx) {
@@ -898,9 +944,13 @@ func (z *erasureServerPools) decommissionInBackground(ctx context.Context, idx i
 		if serverDebugLog {
 			console.Debugln("decommission: currently on bucket", bucket.Name)
 		}
+		stopFn := globalDecommissionMetrics.log(decomMetricDecommissionBucket, idx, bucket.Name)
 		if err := z.decommissionPool(ctx, idx, pool, bucket); err != nil {
+			stopFn(err)
 			return err
 		}
+		stopFn(nil)
+
 		z.poolMetaMutex.Lock()
 		z.poolMeta.BucketDone(idx, bucket)
 		z.poolMeta.save(ctx, z.serverPools)

go.mod

@@ -72,7 +72,6 @@ require (
 	github.com/rs/cors v1.7.0
 	github.com/rs/dnscache v0.0.0-20211102005908-e0241e321417
 	github.com/secure-io/sio-go v0.3.1
-	github.com/shirou/gopsutil v3.21.11+incompatible
 	github.com/shirou/gopsutil/v3 v3.22.6
 	github.com/streadway/amqp v1.0.0
 	github.com/tinylib/msgp v1.1.7-0.20211026165309-e818a1881b0e

go.sum

@@ -804,8 +804,6 @@ github.com/scylladb/termtables v0.0.0-20191203121021-c4c0b6d42ff4/go.mod h1:C1a7
 github.com/sean-/seed v0.0.0-20170313163322-e2103e2c3529/go.mod h1:DxrIzT+xaE7yg65j358z/aeFdxmN0P9QXhEzd20vsDc=
 github.com/secure-io/sio-go v0.3.1 h1:dNvY9awjabXTYGsTF1PiCySl9Ltofk9GA3VdWlo7rRc=
 github.com/secure-io/sio-go v0.3.1/go.mod h1:+xbkjDzPjwh4Axd07pRKSNriS9SCiYksWnZqdnfpQxs=
-github.com/shirou/gopsutil v3.21.11+incompatible h1:+1+c1VGhc88SSonWP6foOcLhvnKlUeu/erjjvaPEYiI=
-github.com/shirou/gopsutil v3.21.11+incompatible/go.mod h1:5b4v6he4MtMOwMlS0TUMTu2PcXUg8+E1lC7eC3UO/RA=
 github.com/shirou/gopsutil/v3 v3.21.6/go.mod h1:JfVbDpIBLVzT8oKbvMg9P3wEIMDDpVn+LwHTKj0ST88=
 github.com/shirou/gopsutil/v3 v3.22.6 h1:FnHOFOh+cYAM0C30P+zysPISzlknLC5Z1G4EAElznfQ=
 github.com/shirou/gopsutil/v3 v3.22.6/go.mod h1:EdIubSnZhbAvBS1yJ7Xi+AShB/hxwLHOMz4MCYz7yMs=