Merge pull request #7714 from filecoin-project/feat/sched-review

Address Scheduler enhancements (#7703) review
filecoin-project · Nov 30, 2021 · 26c9120 · 26c9120
2 parents 73f16f0 + 71329f6
commit 26c9120
Show file tree

Hide file tree

Showing 4 changed files with 10 additions and 7 deletions.
diff --git a/extern/sector-storage/manager_test.go b/extern/sector-storage/manager_test.go
@@ -322,7 +322,7 @@ func TestRestartWorker(t *testing.T) {
 	defer cleanup()
 
 	localTasks := []sealtasks.TaskType{
-		sealtasks.TTAddPiece, sealtasks.TTPreCommit1, sealtasks.TTCommit1, sealtasks.TTFinalize, sealtasks.TTFetch,
+		sealtasks.TTAddPiece, sealtasks.TTFetch,
 	}
 
 	wds := datastore.NewMapDatastore()
@@ -466,7 +466,7 @@ func TestResUse(t *testing.T) {
 	defer cleanup()
 
 	localTasks := []sealtasks.TaskType{
-		sealtasks.TTAddPiece, sealtasks.TTPreCommit1, sealtasks.TTCommit1, sealtasks.TTFinalize, sealtasks.TTFetch,
+		sealtasks.TTAddPiece, sealtasks.TTFetch,
 	}
 
 	wds := datastore.NewMapDatastore()
@@ -524,7 +524,7 @@ func TestResOverride(t *testing.T) {
 	defer cleanup()
 
 	localTasks := []sealtasks.TaskType{
-		sealtasks.TTAddPiece, sealtasks.TTPreCommit1, sealtasks.TTCommit1, sealtasks.TTFinalize, sealtasks.TTFetch,
+		sealtasks.TTAddPiece, sealtasks.TTFetch,
 	}
 
 	wds := datastore.NewMapDatastore()

diff --git a/extern/sector-storage/sched_resources.go b/extern/sector-storage/sched_resources.go
@@ -74,10 +74,12 @@ func (a *activeResources) canHandleRequest(needRes storiface.Resources, wid stor
 
 	vmemNeeded := needRes.MaxMemory + needRes.BaseMinMemory
 	vmemUsed := a.memUsedMax
-	if vmemUsed < res.MemUsed+res.MemSwapUsed {
-		vmemUsed = res.MemUsed + res.MemSwapUsed
+	workerMemoryReserved := res.MemUsed + res.MemSwapUsed // memory used outside lotus-worker (used by the OS, etc.)
+
+	if vmemUsed < workerMemoryReserved {
+		vmemUsed = workerMemoryReserved
 	}
-	vmemAvail := res.MemPhysical + res.MemSwap - vmemUsed
+	vmemAvail := (res.MemPhysical + res.MemSwap) - vmemUsed
 
 	if vmemNeeded > vmemAvail {
 		log.Debugf("sched: not scheduling on worker %s for %s; not enough virtual memory - need: %dM, have %dM available", wid, caller, vmemNeeded/mib, vmemAvail/mib)

diff --git a/extern/sector-storage/storiface/resources.go b/extern/sector-storage/storiface/resources.go
@@ -14,7 +14,7 @@ import (
 
 type Resources struct {
 	MinMemory uint64 `envname:"MIN_MEMORY"` // What Must be in RAM for decent perf
-	MaxMemory uint64 `envname:"MAX_MEMORY"` // Memory required (swap + ram)
+	MaxMemory uint64 `envname:"MAX_MEMORY"` // Memory required (swap + ram; peak memory usage during task execution)
 
 	// GPUUtilization specifes the number of GPUs a task can use
 	GPUUtilization float64 `envname:"GPU_UTILIZATION"`

diff --git a/extern/sector-storage/testworker_test.go b/extern/sector-storage/testworker_test.go
@@ -109,6 +109,7 @@ func (t *testWorker) Info(ctx context.Context) (storiface.WorkerInfo, error) {
 		Resources: storiface.WorkerResources{
 			MemPhysical: res.MinMemory * 3,
 			MemUsed:     res.MinMemory,
+			MemSwapUsed: 0,
 			MemSwap:     0,
 			CPUs:        32,
 			GPUs:        nil,