Improve caching (#19130)

* Remove lock for cached operations.
* Rename "Relax" to `ReturnLastGood`.
* Add `CacheError` to allow caching values even on errors.
* Add NoWait that will return current value with async fetching if within 2xTTL.
* Make benchmark somewhat representative.

```
Before: BenchmarkCache-12       16408370                63.12 ns/op            0 B/op
After:  BenchmarkCache-12       428282187                2.789 ns/op           0 B/op
```

* Remove `storageRESTClient.scanning`. Nonsensical - RPC clients will not have any idea about scanning.
* Always fetch remote diskinfo metrics and cache them. Seems most calls are requesting metrics.
* Do async fetching of usage caches.

cmd/admin-handlers-users.go

@@ -1203,7 +1203,8 @@ func (a adminAPIHandlers) AccountInfoHandler(w http.ResponseWriter, r *http.Requ
 		bucketStorageCache.TTL = 10 * time.Second
 
 		// Rely on older value if usage loading fails from disk.
-		bucketStorageCache.Relax = true
+		bucketStorageCache.ReturnLastGood = true
+		bucketStorageCache.NoWait = true
 		bucketStorageCache.Update = func() (DataUsageInfo, error) {
 			ctx, done := context.WithTimeout(context.Background(), 2*time.Second)
 			defer done()

cmd/bucket-quota.go

@@ -52,7 +52,8 @@ func (sys *BucketQuotaSys) Init(objAPI ObjectLayer) {
 		// does not update the bucket usage values frequently.
 		bucketStorageCache.TTL = 10 * time.Second
 		// Rely on older value if usage loading fails from disk.
-		bucketStorageCache.Relax = true
+		bucketStorageCache.ReturnLastGood = true
+		bucketStorageCache.NoWait = true
 		bucketStorageCache.Update = func() (DataUsageInfo, error) {
 			ctx, done := context.WithTimeout(context.Background(), 2*time.Second)
 			defer done()

cmd/data-usage.go

@@ -81,7 +81,8 @@ func loadPrefixUsageFromBackend(ctx context.Context, objAPI ObjectLayer, bucket
 		prefixUsageCache.TTL = 30 * time.Second
 
 		// No need to fail upon Update() error, fallback to old value.
-		prefixUsageCache.Relax = true
+		prefixUsageCache.ReturnLastGood = true
+		prefixUsageCache.NoWait = true
 		prefixUsageCache.Update = func() (map[string]uint64, error) {
 			m := make(map[string]uint64)
 			for _, pool := range z.serverPools {

cmd/erasure-server-pool.go

@@ -1851,7 +1851,8 @@ func (z *erasureServerPools) ListBuckets(ctx context.Context, opts BucketOptions
 		listBucketsCache.Once.Do(func() {
 			listBucketsCache.TTL = time.Second
 
-			listBucketsCache.Relax = true
+			listBucketsCache.ReturnLastGood = true
+			listBucketsCache.NoWait = true
 			listBucketsCache.Update = func() ([]BucketInfo, error) {
 				ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 				buckets, err = z.s3Peer.ListBuckets(ctx, opts)

cmd/metrics-v2.go

@@ -357,7 +357,7 @@ type MetricsGroupOpts struct {
 func (g *MetricsGroup) RegisterRead(read func(ctx context.Context) []Metric) {
 	g.metricsCache = cachevalue.New[[]Metric]()
 	g.metricsCache.Once.Do(func() {
-		g.metricsCache.Relax = true
+		g.metricsCache.ReturnLastGood = true
 		g.metricsCache.TTL = g.cacheInterval
 		g.metricsCache.Update = func() ([]Metric, error) {
 			if g.metricsGroupOpts.dependGlobalObjectAPI {

cmd/storage-rest-client.go

@@ -31,7 +31,6 @@ import (
 	"strconv"
 	"strings"
 	"sync"
-	"sync/atomic"
 	"time"
 
 	"github.com/minio/madmin-go/v3"
@@ -157,9 +156,6 @@ func toStorageErr(err error) error {
 
 // Abstracts a remote disk.
 type storageRESTClient struct {
-	// Indicate of NSScanner is in progress in this disk
-	scanning int32
-
 	endpoint    Endpoint
 	restClient  *rest.Client
 	gridConn    *grid.Subroute
@@ -236,8 +232,6 @@ func (client *storageRESTClient) Healing() *healingTracker {
 }
 
 func (client *storageRESTClient) NSScanner(ctx context.Context, cache dataUsageCache, updates chan<- dataUsageEntry, scanMode madmin.HealScanMode, _ func() bool) (dataUsageCache, error) {
-	atomic.AddInt32(&client.scanning, 1)
-	defer atomic.AddInt32(&client.scanning, -1)
 	defer xioutil.SafeClose(updates)
 
 	st, err := storageNSScannerRPC.Call(ctx, client.gridConn, &nsScannerOptions{
@@ -310,8 +304,8 @@ func (client *storageRESTClient) DiskInfo(ctx context.Context, opts DiskInfoOpti
 		return info, errDiskNotFound
 	}
 
-	// if metrics was asked, or it was a NoOp we do not need to cache the value.
-	if opts.Metrics || opts.NoOp {
+	// if 'NoOp' we do not cache the value.
+	if opts.NoOp {
 		ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
 		defer cancel()
 
@@ -325,17 +319,17 @@ func (client *storageRESTClient) DiskInfo(ctx context.Context, opts DiskInfoOpti
 		if info.Error != "" {
 			return info, toStorageErr(errors.New(info.Error))
 		}
-		info.Scanning = atomic.LoadInt32(&client.scanning) == 1
 		return info, nil
 	} // In all other cases cache the value upto 1sec.
 
 	client.diskInfoCache.Once.Do(func() {
 		client.diskInfoCache.TTL = time.Second
+		client.diskInfoCache.CacheError = true
 		client.diskInfoCache.Update = func() (info DiskInfo, err error) {
 			ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
 			defer cancel()
 
-			nopts := DiskInfoOptions{DiskID: client.diskID}
+			nopts := DiskInfoOptions{DiskID: client.diskID, Metrics: true}
 			infop, err := storageDiskInfoRPC.Call(ctx, client.gridConn, &nopts)
 			if err != nil {
 				return info, toStorageErr(err)
@@ -348,9 +342,7 @@ func (client *storageRESTClient) DiskInfo(ctx context.Context, opts DiskInfoOpti
 		}
 	})
 
-	info, err = client.diskInfoCache.Get()
-	info.Scanning = atomic.LoadInt32(&client.scanning) == 1
-	return info, err
+	return client.diskInfoCache.Get()
 }
 
 // MakeVolBulk - create multiple volumes in a bulk operation.