Merge 8e0cf4e into 6375c96

Author: quintenp01

ansible/group_vars/all

@@ -209,6 +209,7 @@ invoker:
   heap: "{{ invoker_heap | default('2g') }}"
   arguments: "{{ invoker_arguments | default('') }}"
   userMemory: "{{ invoker_user_memory | default('2048m') }}"
+  userCpus: "{{ invoker_user_cpus | default() }}"
   # Specify if it is allowed to deploy more than 1 invoker on a single machine.
   allowMultipleInstances: "{{ invoker_allow_multiple_instances | default(false) }}"
   # Specify if it should use docker-runc or docker to pause/unpause containers

ansible/roles/invoker/tasks/deploy.yml

@@ -258,6 +258,7 @@
       "CONFIG_whisk_containerFactory_containerArgs_network": "{{ invoker_container_network_name | default('bridge') }}"
       "INVOKER_CONTAINER_POLICY": "{{ invoker_container_policy_name | default()}}"
       "CONFIG_whisk_containerPool_userMemory": "{{ hostvars[groups['invokers'][invoker_index | int]].user_memory | default(invoker.userMemory) }}"
+      "CONFIG_whisk_containerPool_userCpus": "{{ invoker.userCpus | default() }}"
       "CONFIG_whisk_docker_client_parallelRuns": "{{ invoker_parallel_runs | default() }}"
       "CONFIG_whisk_docker_containerFactory_useRunc": "{{ invoker.useRunc | default(false) | lower }}"
       "WHISK_LOGS_DIR": "{{ whisk_logs_dir }}"

common/scala/src/main/scala/org/apache/openwhisk/core/containerpool/ContainerFactory.scala

@@ -25,7 +25,7 @@ import org.apache.openwhisk.spi.Spi
 
 import scala.concurrent.Future
 import scala.concurrent.duration.FiniteDuration
-import scala.math.max
+import scala.math.{max, round}
 
 case class ContainerArgsConfig(network: String,
                                dnsServers: Seq[String] = Seq.empty,
@@ -55,13 +55,15 @@ case class ContainerPoolConfig(userMemory: ByteSize,
                                prewarmPromotion: Boolean,
                                memorySyncInterval: FiniteDuration,
                                batchDeletionSize: Int,
+                               userCpus: Option[Double] = None,
                                prewarmContainerCreationConfig: Option[PrewarmContainerCreationConfig] = None) {
   require(
     concurrentPeekFactor > 0 && concurrentPeekFactor <= 1.0,
     s"concurrentPeekFactor must be > 0 and <= 1.0; was $concurrentPeekFactor")
 
   require(prewarmExpirationCheckInterval.toSeconds > 0, "prewarmExpirationCheckInterval must be > 0")
   require(batchDeletionSize > 0, "batch deletion size must be > 0")
+  require(userCpus.forall(_ > 0), "userCpus must be > 0")
 
   /**
    * The shareFactor indicates the number of containers that would share a single core, on average.
@@ -73,6 +75,16 @@ case class ContainerPoolConfig(userMemory: ByteSize,
   // Grant more CPU to a container if it allocates more memory.
   def cpuShare(reservedMemory: ByteSize) =
     max((totalShare / (userMemory.toBytes / reservedMemory.toBytes)).toInt, 2) // The minimum allowed cpu-shares is 2
+
+  private val minContainerCpus = 0.01 // The minimum cpus allowed by docker is 0.01
+  private val roundingMultiplier = 100000
+  def cpuLimit(reservedMemory: ByteSize): Option[Double] = {
+    userCpus.map(c => {
+      val containerCpus = c / (userMemory.toBytes / reservedMemory.toBytes)
+      val roundedContainerCpus = round(containerCpus * roundingMultiplier).toDouble / roundingMultiplier // Only use decimal precision of 5
+      max(roundedContainerCpus, minContainerCpus)
+    })
+  }
 }
 
 case class PrewarmContainerCreationConfig(maxConcurrent: Int, creationDelay: FiniteDuration) {
@@ -116,16 +128,18 @@ trait ContainerFactory {
     userProvidedImage: Boolean,
     memory: ByteSize,
     cpuShares: Int,
+    cpuLimit: Option[Double],
     action: Option[ExecutableWhiskAction])(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
-    createContainer(tid, name, actionImage, userProvidedImage, memory, cpuShares)
+    createContainer(tid, name, actionImage, userProvidedImage, memory, cpuShares, cpuLimit)
   }
 
   def createContainer(tid: TransactionId,
                       name: String,
                       actionImage: ExecManifest.ImageName,
                       userProvidedImage: Boolean,
                       memory: ByteSize,
-                      cpuShares: Int)(implicit config: WhiskConfig, logging: Logging): Future[Container]
+                      cpuShares: Int,
+                      cpuLimit: Option[Double])(implicit config: WhiskConfig, logging: Logging): Future[Container]
 
   /** perform any initialization */
   def init(): Unit

common/scala/src/main/scala/org/apache/openwhisk/core/yarn/YARNContainerFactory.scala

@@ -123,7 +123,8 @@ class YARNContainerFactory(actorSystem: ActorSystem,
     actionImage: ExecManifest.ImageName,
     unuseduserProvidedImage: Boolean,
     unusedmemory: ByteSize,
-    unusedcpuShares: Int)(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
+    unusedcpuShares: Int,
+    unusedcpuLimit: Option[Double])(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
     implicit val timeout: Timeout = Timeout(containerStartTimeoutMS.milliseconds)
 
     //First send the create command to YARN, then with a different actor, wait for the container to be ready

core/invoker/src/main/resources/application.conf

@@ -71,6 +71,8 @@ whisk {
     prewarm-promotion: false # if true, action can take prewarm container which has bigger memory
     memory-sync-interval: 1 second # period to sync memory info to etcd
     batch-deletion-size: 10 # batch size for removing containers when disable invoker, too big value may cause docker/k8s overload
+    # optional setting to specify the total allocatable cpus for all action containers, each container will get a fraction of this proportional to its allocated memory to limit the cpu
+    # user-cpus: 1
   }
 
   kubernetes {

core/invoker/src/main/scala/org/apache/openwhisk/core/containerpool/ContainerProxy.scala

@@ -250,6 +250,7 @@ class ContainerProxy(factory: (TransactionId,
                                Boolean,
                                ByteSize,
                                Int,
+                               Option[Double],
                                Option[ExecutableWhiskAction]) => Future[Container],
                      sendActiveAck: ActiveAck,
                      storeActivation: (TransactionId, WhiskActivation, Boolean, UserContext) => Future[Any],
@@ -288,6 +289,7 @@ class ContainerProxy(factory: (TransactionId,
         job.exec.pull,
         job.memoryLimit,
         poolConfig.cpuShare(job.memoryLimit),
+        poolConfig.cpuLimit(job.memoryLimit),
         None)
         .map(container =>
           PreWarmCompleted(PreWarmedData(container, job.exec.kind, job.memoryLimit, expires = job.ttl.map(_.fromNow))))
@@ -307,6 +309,7 @@ class ContainerProxy(factory: (TransactionId,
         job.action.exec.pull,
         job.action.limits.memory.megabytes.MB,
         poolConfig.cpuShare(job.action.limits.memory.megabytes.MB),
+        poolConfig.cpuLimit(job.action.limits.memory.megabytes.MB),
         Some(job.action))
 
       // container factory will either yield a new container ready to execute the action, or
@@ -978,6 +981,7 @@ object ContainerProxy {
                       Boolean,
                       ByteSize,
                       Int,
+                      Option[Double],
                       Option[ExecutableWhiskAction]) => Future[Container],
             ack: ActiveAck,
             store: (TransactionId, WhiskActivation, Boolean, UserContext) => Future[Any],

core/invoker/src/main/scala/org/apache/openwhisk/core/containerpool/docker/DockerContainer.scala

@@ -63,6 +63,7 @@ object DockerContainer {
              registryConfig: Option[RuntimesRegistryConfig] = None,
              memory: ByteSize = 256.MB,
              cpuShares: Int = 0,
+             cpuLimit: Option[Double] = None,
              environment: Map[String, String] = Map.empty,
              network: String = "bridge",
              dnsServers: Seq[String] = Seq.empty,
@@ -101,6 +102,7 @@ object DockerContainer {
       dnsSearch.flatMap(d => Seq("--dns-search", d)) ++
       dnsOptions.flatMap(d => Seq(dnsOptString, d)) ++
       name.map(n => Seq("--name", n)).getOrElse(Seq.empty) ++
+      cpuLimit.map(c => Seq("--cpus", c.toString)).getOrElse(Seq.empty) ++
       params
 
     val registryConfigUrl = registryConfig.map(_.url).getOrElse("")

core/invoker/src/main/scala/org/apache/openwhisk/core/containerpool/docker/DockerContainerFactory.scala

@@ -57,12 +57,14 @@ class DockerContainerFactory(instance: InvokerInstanceId,
     extends ContainerFactory {
 
   /** Create a container using docker cli */
-  override def createContainer(tid: TransactionId,
-                               name: String,
-                               actionImage: ExecManifest.ImageName,
-                               userProvidedImage: Boolean,
-                               memory: ByteSize,
-                               cpuShares: Int)(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
+  override def createContainer(
+    tid: TransactionId,
+    name: String,
+    actionImage: ExecManifest.ImageName,
+    userProvidedImage: Boolean,
+    memory: ByteSize,
+    cpuShares: Int,
+    cpuLimit: Option[Double])(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
     val registryConfig =
       ContainerFactory.resolveRegistryConfig(userProvidedImage, runtimesRegistryConfig, userImagesRegistryConfig)
     val image = if (userProvidedImage) Left(actionImage) else Right(actionImage)
@@ -72,6 +74,7 @@ class DockerContainerFactory(instance: InvokerInstanceId,
       registryConfig = Some(registryConfig),
       memory = memory,
       cpuShares = cpuShares,
+      cpuLimit = cpuLimit,
       environment = Map("__OW_API_HOST" -> config.wskApiHost) ++ containerArgsConfig.extraEnvVarMap,
       network = containerArgsConfig.network,
       dnsServers = containerArgsConfig.dnsServers,

core/invoker/src/main/scala/org/apache/openwhisk/core/containerpool/docker/StandaloneDockerContainerFactory.scala

@@ -61,12 +61,14 @@ class StandaloneDockerContainerFactory(instance: InvokerInstanceId, parameters:
   private val pulledImages = new TrieMap[String, Boolean]()
   private val factoryConfig = loadConfigOrThrow[StandaloneDockerConfig](ConfigKeys.standaloneDockerContainerFactory)
 
-  override def createContainer(tid: TransactionId,
-                               name: String,
-                               actionImage: ExecManifest.ImageName,
-                               userProvidedImage: Boolean,
-                               memory: ByteSize,
-                               cpuShares: Int)(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
+  override def createContainer(
+    tid: TransactionId,
+    name: String,
+    actionImage: ExecManifest.ImageName,
+    userProvidedImage: Boolean,
+    memory: ByteSize,
+    cpuShares: Int,
+    cpuLimit: Option[Double])(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
 
     //For standalone server usage we would also want to pull the OpenWhisk provided image so as to ensure if
     //local setup does not have the image then it pulls it down
@@ -84,7 +86,7 @@ class StandaloneDockerContainerFactory(instance: InvokerInstanceId, parameters:
         }
       } else Future.successful(true)
 
-    pulled.flatMap(_ => super.createContainer(tid, name, actionImage, userProvidedImage, memory, cpuShares))
+    pulled.flatMap(_ => super.createContainer(tid, name, actionImage, userProvidedImage, memory, cpuShares, cpuLimit))
   }
 
   override def init(): Unit = {

core/invoker/src/main/scala/org/apache/openwhisk/core/containerpool/kubernetes/KubernetesContainerFactory.scala

@@ -67,12 +67,14 @@ class KubernetesContainerFactory(
     Await.ready(cleaning, KubernetesContainerFactoryProvider.runtimeDeleteTimeout)
   }
 
-  override def createContainer(tid: TransactionId,
-                               name: String,
-                               actionImage: ImageName,
-                               userProvidedImage: Boolean,
-                               memory: ByteSize,
-                               cpuShares: Int)(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
+  override def createContainer(
+    tid: TransactionId,
+    name: String,
+    actionImage: ImageName,
+    userProvidedImage: Boolean,
+    memory: ByteSize,
+    cpuShares: Int,
+    cpuLimit: Option[Double])(implicit config: WhiskConfig, logging: Logging): Future[Container] = {
     val image = actionImage.resolveImageName(Some(
       ContainerFactory.resolveRegistryConfig(userProvidedImage, runtimesRegistryConfig, userImagesRegistryConfig).url))