Merge branch '18842-arv-mount-disk-config' refs #18842

[arvados.git] / lib / dispatchcloud / node_size.go

diff --git a/lib/dispatchcloud/node_size.go b/lib/dispatchcloud/node_size.go
index aa2cd7d5696d46a7184b2edbb39d45b6d4149ead..0b394f4cfe4f76849fc2eb42541ed613e325921f 100644 (file)
--- a/lib/dispatchcloud/node_size.go
+++ b/lib/dispatchcloud/node_size.go
@@ -56,7 +56,7 @@ func estimateDockerImageSize(collectionPDH string) int64 {
 // EstimateScratchSpace estimates how much available disk space (in
 // bytes) is needed to run the container by summing the capacity
 // requested by 'tmp' mounts plus disk space required to load the
-// Docker image.
+// Docker image plus arv-mount block cache.
 func EstimateScratchSpace(ctr *arvados.Container) (needScratch int64) {
        for _, m := range ctr.Mounts {
                if m.Kind == "tmp" {
@@ -80,20 +80,23 @@ func EstimateScratchSpace(ctr *arvados.Container) (needScratch int64) {
        // Now reserve space for the extracted image on disk.
        needScratch += dockerImageSize
 
+       // Now reserve space the arv-mount disk cache
+       needScratch += ctr.RuntimeConstraints.KeepCacheDisk
+
        return
 }
 
-// compareVersion returns true if vs1 >= vs2, otherwise false
-func compareVersion(vs1 string, vs2 string) bool {
+// compareVersion returns true if vs1 < vs2, otherwise false
+func versionLess(vs1 string, vs2 string) (bool, error) {
        v1, err := strconv.ParseFloat(vs1, 64)
        if err != nil {
-               return false
+               return false, err
        }
        v2, err := strconv.ParseFloat(vs2, 64)
        if err != nil {
-               return false
+               return false, err
        }
-       return v1 >= v2
+       return v1 < v2, nil
 }
 
 // ChooseInstanceType returns the cheapest available
@@ -110,11 +113,19 @@ func ChooseInstanceType(cc *arvados.Cluster, ctr *arvados.Container) (best arvad
 
        needRAM := ctr.RuntimeConstraints.RAM + ctr.RuntimeConstraints.KeepCacheRAM
        needRAM += int64(cc.Containers.ReserveExtraRAM)
-       needRAM += int64(cc.Containers.LocalKeepBlobBuffersPerVCPU * needVCPUs * (1 << 26))
+       if cc.Containers.LocalKeepBlobBuffersPerVCPU > 0 {
+               // + 200 MiB for keepstore process + 10% for GOGC=10
+               needRAM += 220 << 20
+               // + 64 MiB for each blob buffer + 10% for GOGC=10
+               needRAM += int64(cc.Containers.LocalKeepBlobBuffersPerVCPU * needVCPUs * (1 << 26) * 11 / 10)
+       }
        needRAM = (needRAM * 100) / int64(100-discountConfiguredRAMPercent)
 
        ok := false
        for _, it := range cc.InstanceTypes {
+               driverInsuff, driverErr := versionLess(it.CUDA.DriverVersion, ctr.RuntimeConstraints.CUDA.DriverVersion)
+               capabilityInsuff, capabilityErr := versionLess(it.CUDA.HardwareCapability, ctr.RuntimeConstraints.CUDA.HardwareCapability)
+
                switch {
                // reasons to reject a node
                case ok && it.Price > best.Price: // already selected a node, and this one is more expensive
@@ -123,9 +134,9 @@ func ChooseInstanceType(cc *arvados.Cluster, ctr *arvados.Container) (best arvad
                case it.VCPUs < needVCPUs: // insufficient VCPUs
                case it.Preemptible != ctr.SchedulingParameters.Preemptible: // wrong preemptable setting
                case it.Price == best.Price && (it.RAM < best.RAM || it.VCPUs < best.VCPUs): // same price, worse specs
-               case it.CUDA.DeviceCount < ctr.RuntimeConstraints.CUDADeviceCount: // insufficient CUDA devices
-               case it.CUDA.DeviceCount > 0 && !compareVersion(it.CUDA.DriverVersion, ctr.RuntimeConstraints.CUDADriverVersion): // insufficient driver version
-               case it.CUDA.DeviceCount > 0 && !compareVersion(it.CUDA.HardwareCapability, ctr.RuntimeConstraints.CUDAHardwareCapability): // insufficient hardware capability
+               case it.CUDA.DeviceCount < ctr.RuntimeConstraints.CUDA.DeviceCount: // insufficient CUDA devices
+               case ctr.RuntimeConstraints.CUDA.DeviceCount > 0 && (driverInsuff || driverErr != nil): // insufficient driver version
+               case ctr.RuntimeConstraints.CUDA.DeviceCount > 0 && (capabilityInsuff || capabilityErr != nil): // insufficient hardware capability
                        // Don't select this node
                default:
                        // Didn't reject the node, so select it