ResourceManager Recovery Properties

Property	Description
yarn.resourcemanager.recovery.enabled	Enables the Resourcemanager to recovery based on the information in the ResourceManager state store. The default, set by configure.sh, is `true`.
yarn.resourcemanager.am.max-attempts	The maximum number of application attempts. This is a global setting for all ApplicationMaster nodes. You can configure an individual maximum number of application attempts for each ApplicationMaster node, but this property sets a global upper bound that overrides the individual node configuration. The default, set in yarn-default.xml, is 2.
mapreduce.am.max-attempts	The maximum number of MapReduce application attempts. If this value is larger than the value set by the ResourceManager, the ResourceManager value will override this value. The default number is set to 2, to allow at least one retry for AM. This property is set in mapred-default.xml.
yarn.resourcemanager.fs.state-store.uri	URI pointing to the location of the FileSystem path where the ResourceManager state is stored. The default value is configured to the path for the ResourceManager volume (`/var/mapr/cluster/yarn/rm/system`). If the FileSystem name is not provided, the system uses the value specified in the `fs.default.name` specified in the `core-site.xml` file.
yarn.resourcemanager.fs.state-store.retry-policy-spec	Specifies the retry policy for the file system client. This policy is specified in pairs of values for the sleep time, in milliseconds, and number of retries. Each pair is enclosed in parentheses, such as `(1000,20), (2000,30)`. The previous example sleeps for 1000 milliseconds for twenty retries, then thirty more retries 2000 milliseconds apart. The default, set in yarn-default.xml, is `(2000,500)`.
yarn.resourcemanager.store.class	The class name of the state-store to be used for saving application/attempt state and the credentials. The available state-store implementations are `org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore`, a ZooKeeper based state-store implementation, and `org.apache.hadoop.yarn.server.resourcemanager.recovery.FileSystemRMStateStore`, a state-store implementation based on file system. The default, yarn-default.xml, is `org.apache.hadoop.yarn.server.resourcemanager.recovery.FileSystemRMStateStore`.
yarn.resourcemanager.state-store.max-completed-applications	The maximum number of completed applications that the state store retains, which is a number less than or equal to `${yarn.resourcemanager.max-completed-applications}`. The default value is 10000. This setting ensures that the applications kept in the state store are consistent with the applications in ResourceManager memory. Any value larger than `${yarn.resourcemanager.max-completed-applications}` is reset to the default. The value of this property affects ResourceManager recovery performance.Typically, a smaller value optimizes performance for recovery.
yarn.resourcemanager.zk-address	A comma-separated list of Host:Port pairs. Each corresponds to a ZooKeeper server, such as 127.0.0.1:5181,127.0.0.1:5181,127.0.0.1:5181. These hosts are used by the ResourceManager to store state.
yarn.resourcemanager.zk-state-store.parent-path	The full path of the root znode where ResourceManager state is stored. The default value is`/rmstore`.
yarn.resourcemanager.zk-num-retries	Number of times the ResourceManager tries to connect to the ZooKeeper server when the connection is lost. The default value is 500.
yarn.resourcemanager.zk-retry-interval-ms	The interval between retries, in milliseconds, when connecting to a ZooKeeper server. The default value is 2000.
yarn.resourcemanager.zk-timeout-ms	The ZooKeeper session timeout in milliseconds. The ZooKeeper server uses this configuration to determine session expiration. Sessions expire when the server does not receive a heartbeat from the client within the session timeout period. The default value is 10000.
yarn.resourcemanager.zk-acl	ACLs that set permissions on ZooKeeper znodes. The default value is `world:anyone:rwcda`

yarn.resourcemanager.recovery.enabled

Enables the Resourcemanager to recovery based on the information in the ResourceManager state store.

The default, set by configure.sh, is true.

yarn.resourcemanager.am.max-attempts

The maximum number of application attempts. This is a global setting for all ApplicationMaster nodes.

You can configure an individual maximum number of application attempts for each ApplicationMaster node, but this property sets a global upper bound that overrides the individual node configuration.

The default, set in yarn-default.xml, is 2.

mapreduce.am.max-attempts

The maximum number of MapReduce application attempts. If this value is larger than the value set by the ResourceManager, the ResourceManager value will override this value. The default number is set to 2, to allow at least one retry for AM. This property is set in mapred-default.xml.

yarn.resourcemanager.fs.state-store.uri

URI pointing to the location of the FileSystem path where the ResourceManager state is stored.

The default value is configured to the path for the ResourceManager volume (/var/mapr/cluster/yarn/rm/system).

If the FileSystem name is not provided, the system uses the value specified in the fs.default.name specified in the core-site.xml file.

yarn.resourcemanager.fs.state-store.retry-policy-spec

Specifies the retry policy for the file system client.

This policy is specified in pairs of values for the sleep time, in milliseconds, and number of retries.

Each pair is enclosed in parentheses, such as (1000,20), (2000,30).

The previous example sleeps for 1000 milliseconds for twenty retries, then thirty more retries 2000 milliseconds apart.

The default, set in yarn-default.xml, is (2000,500).

yarn.resourcemanager.store.class

The class name of the state-store to be used for saving application/attempt state and the credentials.

The available state-store implementations are org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore, a ZooKeeper based state-store implementation, and org.apache.hadoop.yarn.server.resourcemanager.recovery.FileSystemRMStateStore, a state-store implementation based on file system.

The default, yarn-default.xml, is org.apache.hadoop.yarn.server.resourcemanager.recovery.FileSystemRMStateStore.

yarn.resourcemanager.state-store.max-completed-applications

The maximum number of completed applications that the state store retains, which is a number less than or equal to ${yarn.resourcemanager.max-completed-applications}.

The default value is 10000. This setting ensures that the applications kept in the state store are consistent with the applications in ResourceManager memory.

Any value larger than ${yarn.resourcemanager.max-completed-applications} is reset to the default.

The value of this property affects ResourceManager recovery performance.Typically, a smaller value optimizes performance for recovery.

yarn.resourcemanager.zk-address

A comma-separated list of Host:Port pairs. Each corresponds to a ZooKeeper server, such as 127.0.0.1:5181,127.0.0.1:5181,127.0.0.1:5181.

These hosts are used by the ResourceManager to store state.

yarn.resourcemanager.zk-state-store.parent-path

The full path of the root znode where ResourceManager state is stored. The default value is/rmstore.

yarn.resourcemanager.zk-num-retries

Number of times the ResourceManager tries to connect to the ZooKeeper server when the connection is lost.

The default value is 500.

yarn.resourcemanager.zk-retry-interval-ms

The interval between retries, in milliseconds, when connecting to a ZooKeeper server. The default value is 2000.

yarn.resourcemanager.zk-timeout-ms

The ZooKeeper session timeout in milliseconds. The ZooKeeper server uses this configuration to determine session expiration.

Sessions expire when the server does not receive a heartbeat from the client within the session timeout period. The default value is 10000.

yarn.resourcemanager.zk-acl

ACLs that set permissions on ZooKeeper znodes. The default value is world:anyone:rwcda