warden.conf

The file /opt/mapr/conf/warden.conf controls parameters related to MapR services and the warden. Most of the parameters are not intended to be edited directly by users. The following table shows the parameters of interest:

NOTE: When defining heapsize values for services, keep in mind that service.heapsize.percent is bounded by service.heapsize.min, if defined, and service.heapsize.max.

Parameter

Sample Value

Description

service.command.hbmaster.heapsize.percent

4

The percentage of heap space reserved for the HBase Master.

service.command.hbmaster.heapsize.max

512

The maximum heap space, specified in MB, that can be used by the HBase Master.

service.command.hbmaster.heapsize.min

64

The minimum heap space, specified in MB, for use by the HBase Master.

service.command.hbregion.heapsize.percent

25

The percentage of heap space reserved for the HBase Region Server.

service.command.hbregion.heapsize.max

4000

The maximum heap space, specified in MB, that can be used by the HBase Region Server.

service.command.hbregion.heapsize.min

1000

The minimum heap space, specified in MB, for use by the HBase Region Server.

service.command.cldb.heapsize.percent

8

The percentage of heap space reserved for the CLDB.

service.command.cldb.heapsize.max

4000

The maximum heap space, specified in MB, that can be used by the CLDB.

service.command.cldb.heapsize.min

256

The minimum heap space, specified in MB, for use by the CLDB.

service.command.cldb.retryinterval.time.sec

600

Specifies an interval in seconds. The warden attempts to restart a failed CLDB service when this interval expires.

service.command.jt.heapsize.percent

10

The percentage of heap space, specified in MB, reserved for the JobTracker. Memory allocation for JobTracker is only used to calculate total memory required for all services to run. The -Xmx JobTracker value is not set, allowing memory on JobTracker to grow as needed. If an upper limit on memory is strongly desired, set the HADOOP_HEAPSIZE variable in /opt/mapr/hadoop/hadoop-0.20.2/conf/hadoop-env.sh. The value is specified in MB.

service.command.jt.heapsize.max

5000

The maximum heap space, specified in MB, that can be used by the JobTracker. Memory allocation for JobTracker is only used to calculate total memory required for all services to run. The -Xmx JobTracker value is not set, allowing memory on JobTracker to grow as needed. If an upper limit on memory is strongly desired, set the HADOOP_HEAPSIZE variable in /opt/mapr/hadoop/hadoop-0.20.2/conf/hadoop-env.sh.

service.command.jt.heapsize.min

256

The minimum heap space, specified in MB, for use by the JobTracker. Memory allocation for JobTracker is only used to calculate total memory required for all services to run. The -Xmx JobTracker value is not set, allowing memory on JobTracker to grow as needed. If an upper limit on memory is strongly desired, set the HADOOP_HEAPSIZE variable in /opt/mapr/hadoop/hadoop-0.20.2/conf/hadoop-env.sh.

service.command.mfs.heapsize.percent

35

The percentage of heap space reserved for the MapR-FS. Restart the Warden after modifying this setting.
service.command.mfs.heapsize.maxpercent

85

 The maximum percentage of heap space that can be allocated to the MapR-FS. Restart the Warden after modifying this setting.

service.command.mfs.heapsize.min

512

The minimum heap space, specified in MB, that can be allocated to the MapR-FS. Restart the Warden after modifying this setting.

service.command.tt.heapsize.percent

2

The percentage of heap space reserved for the TaskTracker. Memory allocation for TaskTracker is only used to calculate total memory required for all services to run. The -Xmx TaskTracker value is not set, allowing memory on TaskTracker to grow as needed. If an upper limit on memory is strongly desired, set the HADOOP_HEAPSIZE variable in /opt/mapr/hadoop/hadoop-0.20.2/conf/hadoop-env.sh.

service.command.tt.heapsize.max

325

The maximum heap space, specified in MB, that can be used by the TaskTracker. Memory allocation for TaskTracker is only used to calculate total memory required for all services to run. The -Xmx TaskTracker value is not set, allowing memory on TaskTracker to grow as needed. If an upper limit on memory is strongly desired, set the HADOOP_HEAPSIZE variable in /opt/mapr/hadoop/hadoop-0.20.2/conf/hadoop-env.sh.

service.command.tt.heapsize.min

64

The minimum heap space, specified in MB, for use by the TaskTracker. Memory allocation for TaskTracker is only used to calculate total memory required for all services to run. The -Xmx TaskTracker value is not set, allowing memory on TaskTracker to grow as needed. If an upper limit on memory is strongly desired, set the HADOOP_HEAPSIZE variable in /opt/mapr/hadoop/hadoop-0.20.2/conf/hadoop-env.sh.

service.command.webserver.heapsize.percent

3

The percentage of heap space reserved for the MapR Control System.

service.command.webserver.heapsize.max

750

The maximum heap space, specified in MB, that can be used by the MapR Control System.

service.command.webserver.heapsize.min

512

The minimum heap space, specified in MB, for use by the MapR Control System.

service.command.nfs.heapsize.percent 3 The percentage of heap space reserved for the NFS.
service.command.nfs.heapsize.min 64 The minimum heap space, specified in MB, that can be used by the NFS.
service.command.nfs.heapsize.max 1000 The maximum heap space,specified in MB, that can be used by the NFS.

service.command.os.heapsize.percent

3

The percentage of heap space reserved for the operating system.

service.command.os.heapsize.max

750

The maximum heap space, specified in MB, that can be used by the operating system.

service.command.os.heapsize.min

256

The minimum heap space, specified in MB, for use by the operating system.

service.nice.value

-10

The nice priority under which all services will run.

zookeeper.servers

10.250.1.61:5181

The list of ZooKeeper servers.
services.memoryallocation.alarm.threshold 95 The maximum amount of system memory that services running on the node can use before triggering the NODE_ALARM_MEMORY_ALLOCATION_EXCEEDED alarm.

services.retries

3

The number of times the Warden tries to restart a service that fails.

services.resetretries.time.sec

3600

Specifies a time interval in seconds. The services.retries parameter sets the number of times that the warden attempts to restart failing services within this interval.

services.retryinterval.time.sec

1800

The number of seconds after which the warden will again attempt several times to start a failed service. The number of attempts after each interval is specified by the parameter services.retries.

cldb.port

7222

The port for communicating with the CLDB.

mfs.port

5660

The port for communicating with the FileServer.

hbmaster.port

60000

The port for communicating with the HBase Master.

hoststats.port

5660

The port for communicating with the HostStats service.
jt.port

9001

The port for communicating with the JobTracker.

jt.response.timeout.minutes

10

Specifies an interval in minutes. The warden kills JobTracker services that do not respond within the specified interval and restarts them as normal for failed services.

kvstore.port

5660

The port for communicating with the Key/Value Store.

mapr.home.dir

/opt/mapr

The directory where MapR is installed.

centralconfig.enabled

true

Specifies whether to enable central configuration.

pollcentralconfig.interval.seconds

300

How often to check for configuration updates, in seconds.

rpc.drop

false

Drop outstanding metrics when the queue to send to hoststats is too large.

hs.rpcon

true

Whether or not to configure Job Management.

hs.port

1111

Hoststats listening port for Metrics RPC activity.

hs.host

localhost

Hoststats hostname for RPC activity.

log.retention.time

864000000

All .log and .out files in the cluster are kept for a time period defined by the value of the log.retention.time parameter in milliseconds. The default value is ten days. Restart the Warden after changing this value.

log.retention.exceptions

Retains the following log files instead of removing them during the log file cleanup that occurs every ten days: cldb.log,hoststats.log,configure.log,mfs.log-*

You can modify the list. Partial file names and asterisks are accepted. Log files listed as exceptions are retained indefinitely.

enable.overcommit

false

Set this value to true to allow services to start up even if their memory demands exceed the memory provided by the node.
mr1.memory.percent 50 The percentage of memory allocated to MapReduce v1 jobs. The remaining memory is allocated to applications. For more information, see Resource Allocation for Jobs and Applications.
mr1.cpu.percent 50 The percentage of CPU allocated to MapReduce v1 jobs. The remaining memory is allocated to applications. For more information, see Resource Allocation for Jobs and Applications.
mr1.disk.percent 50 The percentage of memory allocated to MapReduce v1 jobs. The remaining memory is allocated to applications. For more information, see Resource Allocation for Jobs and Applications.
isDB true Specifies if MapR-DB is in use. When this value is false, the service.command.mfs.heapsize.percent is set to 20. Do not manually edit this value. For more information, see Memory Allocation for Nodes.

For information on configuration files for additional services, see warden.<servicename>.conf.