Use concurrency in poolmanager as per old behaviour (#2876)

* Use concurrency in poolmanager as per old behaviour * Update code comments --------- Signed-off-by: Sanket Sudake <sanketsudake@gmail.com>
fission · Nov 21, 2023 · b85ba9e · b85ba9e
1 parent d23ed57
commit b85ba9e
Show file tree

Hide file tree

Showing 2 changed files with 33 additions and 15 deletions.
diff --git a/pkg/executor/fscache/poolcache.go b/pkg/executor/fscache/poolcache.go
@@ -96,7 +96,6 @@ type (
 )
 
 // NewPoolCache create a Cache object
-
 func NewPoolCache(logger *zap.Logger) *PoolCache {
 	c := &PoolCache{
 		cache:          make(map[crd.CacheKeyURG]*funcSvcGroup),
@@ -122,6 +121,7 @@ func (c *PoolCache) service() {
 		case getValue:
 			funcSvcGroup, ok := c.cache[req.function]
 			if !ok {
+				// first request for this function, create a new group
 				c.cache[req.function] = NewFuncSvcGroup()
 				c.cache[req.function].svcWaiting++
 				resp.error = ferror.MakeError(ferror.ErrorNotFound,
@@ -131,6 +131,7 @@ func (c *PoolCache) service() {
 			}
 			found := false
 			totalActiveRequests := 0
+			// check if any specialized pod is available
 			for addr := range funcSvcGroup.svcs {
 				totalActiveRequests += funcSvcGroup.svcs[addr].activeRequests
 				if funcSvcGroup.svcs[addr].activeRequests < req.requestsPerPod &&
@@ -145,12 +146,22 @@ func (c *PoolCache) service() {
 					break
 				}
 			}
+			// if specialized pod is available then return svc
 			if found {
 				req.responseChannel <- resp
 				continue
 			}
-			specializationInProgress := funcSvcGroup.svcWaiting - funcSvcGroup.queue.Len()
-			capacity := ((specializationInProgress + len(funcSvcGroup.svcs)) * req.requestsPerPod) - (totalActiveRequests + funcSvcGroup.svcWaiting)
+			concurrencyUsed := len(funcSvcGroup.svcs) + (funcSvcGroup.svcWaiting - funcSvcGroup.queue.Len())
+			// if concurrency is available then be aggressive and use it as we are not sure if specialization will complete for other requests
+			if req.concurrency > 0 && concurrencyUsed < req.concurrency {
+				funcSvcGroup.svcWaiting++
+				resp.error = ferror.MakeError(ferror.ErrorNotFound, fmt.Sprintf("function '%s' not found", req.function))
+				req.responseChannel <- resp
+				continue
+			}
+			// if no concurrency is available then check if there is any virtual capacity in the existing pods to serve the request in future
+			// if specialization doesnt complete within request then request will be timeout
+			capacity := (concurrencyUsed * req.requestsPerPod) - (totalActiveRequests + funcSvcGroup.svcWaiting)
 			if capacity > 0 {
 				funcSvcGroup.svcWaiting++
 				svcWait := &svcWait{
@@ -164,8 +175,8 @@ func (c *PoolCache) service() {
 			}
 
 			// concurrency should not be set to zero and
-			//sum of specialization in progress and specialized pods should be less then req.concurrency
-			if req.concurrency > 0 && (specializationInProgress+len(funcSvcGroup.svcs)) >= req.concurrency {
+			// sum of specialization in progress and specialized pods should be less then req.concurrency
+			if req.concurrency > 0 && concurrencyUsed >= req.concurrency {
 				resp.error = ferror.MakeError(ferror.ErrorTooManyRequests, fmt.Sprintf("function '%s' concurrency '%d' limit reached.", req.function, req.concurrency))
 			} else {
 				funcSvcGroup.svcWaiting++

diff --git a/pkg/executor/fscache/poolcache_test.go b/pkg/executor/fscache/poolcache_test.go
@@ -224,17 +224,19 @@ func TestPoolCacheRequests(t *testing.T) {
 				simultaneous = 1
 			}
 			for i := 1; i <= tt.requests; i++ {
+				reqno := i
 				wg.Add(1)
 				go func(reqno int) {
 					defer wg.Done()
 					svc, err := p.GetSvcValue(context.Background(), key, tt.rpp, tt.concurrency)
 					if err != nil {
 						code, _ := ferror.GetHTTPError(err)
 						if code == http.StatusNotFound {
-							p.SetSvcValue(context.Background(), key, fmt.Sprintf("svc-%d", svcCounter), &FuncSvc{
-								Name: "value",
-							}, resource.MustParse("45m"), tt.rpp, tt.retainPods)
 							atomic.AddUint64(&svcCounter, 1)
+							address := fmt.Sprintf("svc-%d", atomic.LoadUint64(&svcCounter))
+							p.SetSvcValue(context.Background(), key, address, &FuncSvc{
+								Name: address,
+							}, resource.MustParse("45m"), tt.rpp, tt.retainPods)
 						} else {
 							t.Log(reqno, "=>", err)
 							atomic.AddUint64(&failedRequests, 1)
@@ -244,9 +246,12 @@ func TestPoolCacheRequests(t *testing.T) {
 							t.Log(reqno, "=>", "svc is nil")
 							atomic.AddUint64(&failedRequests, 1)
 						}
+						// } else {
+						// 	t.Log(reqno, "=>", svc.Name)
+						// }
 					}
-				}(i)
-				if i%simultaneous == 0 {
+				}(reqno)
+				if reqno%simultaneous == 0 {
 					wg.Wait()
 				}
 			}
@@ -258,10 +263,11 @@ func TestPoolCacheRequests(t *testing.T) {
 			for i := 0; i < tt.concurrency; i++ {
 				for j := 0; j < tt.rpp; j++ {
 					wg.Add(1)
-					go func(i int) {
+					svcno := i
+					go func(svcno int) {
 						defer wg.Done()
-						p.MarkAvailable(key, fmt.Sprintf("svc-%d", i))
-					}(i)
+						p.MarkAvailable(key, fmt.Sprintf("svc-%d", svcno+1))
+					}(svcno)
 				}
 			}
 			wg.Wait()
@@ -270,8 +276,9 @@ func TestPoolCacheRequests(t *testing.T) {
 					UID:        "func",
 					Generation: 2,
 				}
-				p.SetSvcValue(context.Background(), newKey, fmt.Sprintf("svc-%d", svcCounter), &FuncSvc{
-					Name: "value",
+				address := fmt.Sprintf("svc-%d", svcCounter)
+				p.SetSvcValue(context.Background(), newKey, address, &FuncSvc{
+					Name: address,
 				}, resource.MustParse("45m"), tt.rpp, tt.retainPods)
 				funcSvc := p.ListAvailableValue()
 				require.Equal(t, tt.concurrency, len(funcSvc))