Using Cgroups with YARN

Cgroups is a mechanism for aggregating/partitioning sets of tasks, and all their future children, into hierarchical groups with specialized behaviour. Cgroups (v1) is a Linux kernel feature and was merged into kernel version 2.6.24, while Control Group v2 is available since the kernel version 4.5. From a YARN perspective, this allows containers to be limited in their resource usage. A good example of this is CPU usage. Without cgroups, it becomes hard to limit container CPU usage.

Cgroups Configuration

This section describes the configuration variables for using cgroups.

The following settings are related to setting up cgroups. These need to be set in yarn-site.xml.

Configuration Name Description
yarn.nodemanager.container-executor.class This should be set to “org.apache.hadoop.yarn.server.nodemanager.LinuxContainerExecutor”. Cgroups is a Linux kernel feature and is exposed via the LinuxContainerExecutor.
yarn.nodemanager.linux-container-executor.resources-handler.class This should be set to “org.apache.hadoop.yarn.server.nodemanager.util.CgroupsLCEResourcesHandler”. Using the LinuxContainerExecutor doesn’t force you to use cgroups. If you wish to use cgroups, the resource-handler-class must be set to CGroupsLCEResourceHandler. DefaultLCEResourcesHandler won’t work.
yarn.nodemanager.linux-container-executor.cgroups.v2.enabled A property to enable cgroup v2 support. Setting this to true YARN will try to use the cgroup v2 structure and controllers. If this setting is true, but no unified (v2) hierarchy is mounted it will automatically fall back to v1. Defaults to false.
yarn.nodemanager.linux-container-executor.cgroups.hierarchy The cgroups hierarchy under which to place YARN proccesses (cannot contain commas). If yarn.nodemanager.linux-container-executor.cgroups.mount is false (that is, if cgroups have been pre-configured) and the YARN user has write access to the parent directory, then the directory will be created. If the directory already exists, the administrator has to give YARN write permissions to it recursively.
yarn.nodemanager.linux-container-executor.cgroups.mount Whether the LCE should attempt to mount cgroups if not found - can be true or false. Mounting is not supported with cgroup v2.
yarn.nodemanager.linux-container-executor.cgroups.mount-path Optional. Where cgroup is located. LCE will try to mount them here, if yarn.nodemanager.linux-container-executor.cgroups.mount is true (and cgroup v1 is used). LCE will try to use cgroups from this location, if yarn.nodemanager.linux-container-executor.cgroups.mount is false. If specified, this path and its subdirectories (cgroup hierarchies) must exist and they should be readable and writable by YARN before the NodeManager is launched. See Cgroups mount options below for details.
yarn.nodemanager.linux-container-executor.cgroups.v2.mount-path Optional. Where cgroup v2 is located. This property needs to be specified only if both cgroup v1 and v2 is used. For example in mixed mode cgroup v1 controllers can be mounted under /sys/fs/cgroup/ (i.e. /sys/fs/cgroup/cpu,cpuacct), while v2 can be mounted in /sys/fs/cgroup/unified folder. If specified, this path (cgroup v2 hierarchy) must exist and it should be readable and writable by YARN before the NodeManager is launched.
yarn.nodemanager.linux-container-executor.group The Unix group of the NodeManager. It should match the setting in “container-executor.cfg”. This configuration is required for validating the secure access of the container-executor binary.

Once cgroup is enabled, the following settings related to limiting resource usage of YARN containers can works:

Configuration Name Description
yarn.nodemanager.resource.percentage-physical-cpu-limit This setting lets you limit the cpu usage of all YARN containers. It sets a hard upper limit on the cumulative CPU usage of the containers. For example, if set to 60, the combined CPU usage of all YARN containers will not exceed 60%.
yarn.nodemanager.linux-container-executor.cgroups.strict-resource-usage Cgroups allows cpu usage limits to be hard or soft. When this setting is true, containers cannot use more CPU usage than allocated even if spare CPU is available. This ensures that containers can only use CPU that they were allocated. When set to false, containers can use spare CPU if available. It should be noted that irrespective of whether set to true or false, at no time can the combined CPU usage of all containers exceed the value specified in “yarn.nodemanager.resource.percentage-physical-cpu-limit”.

Cgroups mount options

YARN uses cgroups through a directory structure mounted into the file system by the kernel. There are three options to attach to cgroups.

Option Description
Discover cgroups mounted already This should be used on newer systems like RHEL7 or Ubuntu16 or if the administrator mounts cgroups before YARN starts. Set yarn.nodemanager.linux-container-executor.cgroups.mount to false and leave other settings set to their defaults. YARN will locate the mount points in /proc/mounts. Common locations include /sys/fs/cgroup and /cgroup. The default location can vary depending on the Linux distribution in use.
Cgroups mounted by YARN IMPORTANT: This option is deprecated due to security reasons with the container-executor.cfg option feature.mount-cgroup.enabled=0 by default. Please mount cgroups before launching YARN.
Cgroups mounted already or linked but not in /proc/mounts If cgroups is accessible through lxcfs or simulated by another filesystem, then point yarn.nodemanager.linux-container-executor.cgroups.mount-path to your cgroups root directory. Set yarn.nodemanager.linux-container-executor.cgroups.mount to false. YARN tries to use this path first, before any cgroup mount point discovery. The path should have a subdirectory in cgroup v1 for each cgroup hierarchy named by the comma separated cgroup subsystems supported like <path>/cpu,cpuacct. Valid subsystem names are cpu, cpuacct, cpuset, memory, net_cls, blkio, freezer, devices.

Cgroups and security

Cgroups itself has no requirements related to security. However, the LinuxContainerExecutor does have some requirements. If running in non-secure mode, by default, the LCE runs all jobs as user “nobody”. This user can be changed by setting “yarn.nodemanager.linux-container-executor.nonsecure-mode.local-user” to the desired user. However, it can also be configured to run jobs as the user submitting the job. In that case “yarn.nodemanager.linux-container-executor.nonsecure-mode.limit-users” should be set to false.

yarn.nodemanager.linux-container-executor.nonsecure-mode.local-user yarn.nodemanager.linux-container-executor.nonsecure-mode.limit-users User running jobs
(default) (default) nobody
yarn (default) yarn
yarn false (User submitting the job)