This is the documentation for Cloudera Manager 5.0.x. Documentation for other versions is available at Cloudera Documentation.

MapReduce

Role groups:

Failover Controller Default Group

Advanced

Display Name Description Related Name Default Value Unit API Name Required
Java Configuration Options for Failover Controller These arguments will be passed as part of the Java command line. Commonly, garbage collection flags or extra debugging flags would be passed here. failover_controller_java_opts false
Failover Controller Advanced Configuration Snippet (Safety Valve) for mapred-site.xml For advanced use only, a string to be inserted into mapred-site.xml for this role only. fc_config_safety_valve false
Failover Controller Logging Advanced Configuration Snippet (Safety Valve) For advanced use only, a string to be inserted into log4j.properties for this role only. log4j_safety_valve false
Heap Dump Directory Path to directory where heap dumps are generated when java.lang.OutOfMemoryError error is thrown. This directory is automatically created if it doesn't exist. However, if this directory already exists, role user must have write access to this directory. If this directory is shared amongst multiple roles, it should have 1777 permissions. Note that the heap dump files are created with 600 permissions and are owned by the role user. The amount of free space in this directory should be greater than the maximum Java Process heap size configured for this role. /tmp oom_heap_dump_dir false
Dump Heap When Out of Memory When set, generates heap dump file when java.lang.OutOfMemoryError is thrown. false oom_heap_dump_enabled true
Kill When Out of Memory When set, a SIGKILL signal is sent to the role process when java.lang.OutOfMemoryError is thrown. true oom_sigkill_enabled true
Automatically Restart Process When set, this role's process is automatically (and transparently) restarted in the event of an unexpected failure. false process_auto_restart true

Logs

Display Name Description Related Name Default Value Unit API Name Required
Failover Controller Log Directory Directory where Failover Controller will place its log files. /var/log/hadoop-0.20-mapreduce failover_controller_log_dir false
Failover Controller Logging Threshold The minimum log level for Failover Controller logs INFO log_threshold false
Failover Controller Maximum Log File Backups The maximum number of rolled log files to keep for Failover Controller logs. Typically used by log4j. 10 max_log_backup_index false
Failover Controller Max Log Size The maximum size, in megabytes, per log file for Failover Controller logs. Typically used by log4j. 200 MiB max_log_size false

Monitoring

Display Name Description Related Name Default Value Unit API Name Required
Enable Health Alerts for this Role When set, Cloudera Manager will send alerts when the health of this role reaches the threshold specified by the EventServer setting eventserver_health_events_alert_threshold true enable_alerts false
Enable Configuration Change Alerts When set, Cloudera Manager will send alerts when this entity's configuration changes. false enable_config_alerts false
File Descriptor Monitoring Thresholds The health test thresholds of the number of file descriptors used. Specified as a percentage of file descriptor limit. Warning: 50.0 %, Critical: 70.0 % failovercontroller_fd_thresholds false
Failover Controller Host Health Test When computing the overall Failover Controller health, consider the host's health. true failovercontroller_host_health_enabled false
Failover Controller Process Health Test Enables the health test that the Failover Controller's process state is consistent with the role configuration true failovercontroller_scm_health_enabled false
Log Directory Free Space Monitoring Absolute Thresholds The health test thresholds for monitoring of free space on the filesystem that contains this role's log directory. Warning: 10 GiB, Critical: 5 GiB log_directory_free_space_absolute_thresholds false
Log Directory Free Space Monitoring Percentage Thresholds The health test thresholds for monitoring of free space on the filesystem that contains this role's log directory. Specified as a percentage of the capacity on that filesystem. This setting is not used if a Log Directory Free Space Monitoring Absolute Thresholds setting is configured. Warning: Never, Critical: Never log_directory_free_space_percentage_thresholds false
Rules to Extract Events from Log Files This file contains the rules which govern how log messages are turned into events by the custom log4j appender that this role loads. It is in JSON format, and is composed of a list of rules. Every log message is evaluated against each of these rules in turn to decide whether or not to send an event for that message. Each rule has some or all of the following fields:
  • alert - whether or not events generated from this rule should be promoted to alerts. A value of "true" will cause alerts to be generated. If not specified, the default is "false".
  • rate (mandatory) - the maximum number of log messages matching this rule that may be sent as events every minute. If more than rate matching log messages are received in a single minute, the extra messages are ignored. If rate is less than 0, the number of messages per minute is unlimited.
  • periodminutes - the number of minutes during which the publisher will only publish rate events or fewer. If not specified, the default is one minute
  • threshold - apply this rule only to messages with this log4j severity level or above. An example is "WARN" for warning level messages or higher.
  • content - match only those messages whose contents match this regular expression.
  • exceptiontype - match only those messages which are part of an exception message. The exception type must match this regular expression.
Example:{"alert": false, "rate": 10, "exceptiontype": "java.lang.StringIndexOutOfBoundsException"} This rule will send events to Cloudera Manager for every StringIndexOutOfBoundsException, up to a maximum of 10 every minute. 		version: 0, rules: [ alert: false, rate: 1, periodminutes: 1, threshold:FATAL, alert: false, rate: 1, periodminutes: 2, exceptiontype: .*, alert: false, rate: 1, periodminutes: 1, threshold:WARN ] log_event_whitelist false
Role Triggers The configured triggers for this role. This is a JSON formatted list of triggers. These triggers are evaluated as part as the health system. Every trigger expression is parsed, and if the trigger condition is met, the list of actions provided in the trigger expression is executed. Each trigger has all of the following fields:
  • triggerName (mandatory) - the name of the trigger. This value must be unique for the specific role.
  • triggerExpression (mandatory) - a tsquery expression representing the trigger.
  • streamThreshold (optional) - the maximum number of streams that can satisfy a condition of a trigger before the condition fires. By default set to 0, and any stream returned will cause the condition to fire.
  • enabled (optional) - by default set to 'true'. If set to 'false' the trigger will not be evaluated.
For example, here is a JSON formatted trigger configured for a DataNode that fires if the DataNode has more than 1500 file-descriptors opened: 
[{"triggerName": "sample-trigger",
  "triggerExpression": "IF (SELECT fd_open WHERE roleName=$ROLENAME and last(fd_open) > 1500) DO health:red",
  "streamThreshold": 0}, "enabled": "true"]
Consult the trigger rules documentation for more details on how to write triggers using tsquery. The JSON format is evolving and may change in the future and as a result backward compatibility is not guaranteed between releases at this time. 		[] role_triggers true
Unexpected Exits Thresholds The health test thresholds for unexpected exits encountered within a recent period specified by the unexpected_exits_window configuration for the role. Warning: Never, Critical: Any unexpected_exits_thresholds false
Unexpected Exits Monitoring Period The period to review when computing unexpected exits. 5 minute(s) unexpected_exits_window false

Performance

Display Name Description Related Name Default Value Unit API Name Required
Maximum Process File Descriptors If configured, overrides the process soft and hard rlimits (also called ulimits) for file descriptors to the configured value. rlimit_fds false

Ports and Addresses

Display Name Description Related Name Default Value Unit API Name Required
Failover controller port The ZooKeeper failover controller port. mapred.ha.zkfc.port 8018 mapred_ha_zkfc_port false

Resource Management

Display Name Description Related Name Default Value Unit API Name Required
Java Heap Size of Failover Controller in Bytes Maximum size for the Java Process heap memory. Passed to Java -Xmx. Measured in bytes. 256 MiB failover_controller_java_heapsize false
Cgroup CPU Shares Number of CPU shares to assign to this role. The greater the number of shares, the larger the share of the host's CPUs that will be given to this role when the host experiences CPU contention. Must be between 2 and 262144. Defaults to 1024 for processes not managed by Cloudera Manager. cpu.shares 1024 rm_cpu_shares true
Cgroup I/O Weight Weight for the read I/O requests issued by this role. The greater the weight, the higher the priority of the requests when the host experiences I/O contention. Must be between 100 and 1000. Defaults to 1000 for processes not managed by Cloudera Manager. blkio.weight 500 rm_io_weight true
Cgroup Memory Hard Limit Hard memory limit to assign to this role, enforced by the Linux kernel. When the limit is reached, the kernel will reclaim pages charged to the process. If reclaiming fails, the kernel may kill the process. Both anonymous as well as page cache pages contribute to the limit. Use a value of -1 B to specify no limit. By default processes not managed by Cloudera Manager will have no limit. memory.limit_in_bytes -1 MiB rm_memory_hard_limit true
Cgroup Memory Soft Limit Soft memory limit to assign to this role, enforced by the Linux kernel. When the limit is reached, the kernel will reclaim pages charged to the process if and only if the host is facing memory pressure. If reclaiming fails, the kernel may kill the process. Both anonymous as well as page cache pages contribute to the limit. Use a value of -1 B to specify no limit. By default processes not managed by Cloudera Manager will have no limit. memory.soft_limit_in_bytes -1 MiB rm_memory_soft_limit true

Gateway Default Group

Advanced

Display Name Description Related Name Default Value Unit API Name Required
MapReduce Client Advanced Configuration Snippet (Safety Valve) for mapred-site.xml For advanced use only, a string to be inserted into the client configuration for mapred-site.xml. mapreduce_client_config_safety_valve false
MapReduce Client Environment Advanced Configuration Snippet for hadoop-env.sh (Safety Valve) For advanced use only, key-value pairs (one on each line) to be inserted into the client configuration for hadoop-env.sh. mapreduce_client_env_safety_valve false
Client Java Configuration Options These are Java command line arguments. Commonly, garbage collection flags or extra debugging flags would be passed here. -Djava.net.preferIPv4Stack=true mapreduce_client_java_opts false

Compression

Display Name Description Related Name Default Value Unit API Name Required
Use Compression on Map Outputs If enabled, uses compression on the map outputs before they are sent across the network. Will be part of generated client configuration. mapred.compress.map.output true mapred_compress_map_output false
Compression Codec of MapReduce Map Output For MapReduce map outputs that are compressed, specify the compression codec to use. Will be part of generated client configuration. mapred.map.output.compression.codec org.apache.hadoop.io.compress.SnappyCodec mapred_map_output_compression_codec false
Compress MapReduce Job Output Compress the output of MapReduce jobs. Will be part of generated client configuration. mapred.output.compress false mapred_output_compress false
Compression Codec of MapReduce Job Output For MapReduce job outputs that are compressed, specify the compression codec to use. Will be part of generated client configuration. mapred.output.compression.codec org.apache.hadoop.io.compress.DefaultCodec mapred_output_compression_codec false
Compression Type of MapReduce Job Output For MapReduce job outputs that are compressed as SequenceFiles, you can select one of these compression type options: NONE, RECORD or BLOCK. Cloudera recommends BLOCK. Will be part of generated client configuration. mapred.output.compression.type BLOCK mapred_output_compression_type false
Compression Level of Codecs Compression level for the codec used to compress MapReduce outputs. Default compression is a balance between speed and compression ratio. zlib.compress.level DEFAULT_COMPRESSION zlib_compress_level false

Jobs

Display Name Description Related Name Default Value Unit API Name Required
Number of Tasks to Run per JVM Number of tasks to run per JVM. If set to -1, there is no limit. Will be part of generated client configuration. mapred.job.reuse.jvm.num.tasks 1 mapred_job_reuse_jvm_num_tasks false
Map Tasks Speculative Execution If enabled, multiple instances of some map tasks may be executed in parallel. mapred.map.tasks.speculative.execution false mapred_map_tasks_speculative_execution false
Number of Map Tasks to Complete Before Reduce Tasks Fraction of the number of map tasks in the job which should be completed before reduce tasks are scheduled for the job. mapred.reduce.slowstart.completed.maps 0.8 mapred_reduce_slowstart_completed_maps false
Default Number of Reduce Tasks per Job The default number of reduce tasks per job. Will be part of generated client configuration. mapred.reduce.tasks 1 mapred_reduce_tasks false
Reduce Tasks Speculative Execution If enabled, multiple instances of some reduce tasks may be executed in parallel. mapred.reduce.tasks.speculative.execution false mapred_reduce_tasks_speculative_execution false
Maximum Time to Retain User Logs The maximum time, in hours, to retain the user logs after job completion. mapred.userlog.retain.hours 1 day(s) mapred_userlog_retain_hours false

Monitoring

Display Name Description Related Name Default Value Unit API Name Required
Enable Configuration Change Alerts When set, Cloudera Manager will send alerts when this entity's configuration changes. false enable_config_alerts false

Other

Display Name Description Related Name Default Value Unit API Name Required
Alternatives Priority The priority level that the client configuration will have in the Alternatives system on the hosts. Higher priority levels will cause Alternatives to prefer this configuration over any others. 92 client_config_priority true
Mapreduce Submit Replication The replication level for submitted job files. mapred.submit.replication 10 mapred_submit_replication false
Mapreduce Task Timeout The number of milliseconds before a task will be terminated if it neither reads an input, writes an output, nor updates its status string. mapred.task.timeout 10 minute(s) mapred_task_timeout false

Performance

Display Name Description Related Name Default Value Unit API Name Required
I/O Sort Factor The number of streams to merge at the same time while sorting files. That is, the number of sort heads to use during the merge sort on the reducer side. This determines the number of open file handles. Merging more files in parallel reduces merge sort iterations and improves run time by eliminating disk I/O. Note that merging more files in parallel uses more memory. If 'io.sort.factor' is set too high or the maximum JVM heap is set too low, excessive garbage collection will occur. The Hadoop default is 10, but Cloudera recommends a higher value. Will be part of generated client configuration. io.sort.factor 64 io_sort_factor false
I/O Sort Memory Buffer (MiB) The total amount of memory buffer, in megabytes, to use while sorting files. Note that this memory comes out of the user JVM heap size (meaning total user JVM heap - this amount of memory = total user usable heap space. Note that Cloudera's default differs from Hadoop's default; Cloudera uses a bigger buffer by default because modern machines often have more RAM. The smallest value across all TaskTrackers will be part of generated client configuration. io.sort.mb 256 MiB io_sort_mb false
I/O Sort Record Percent The percentage of 'io.sort.mb' dedicated to tracking record boundaries. If this value is represented as 'r', and 'io.sort.mb' is represented as 'x', then the maximum number of records collected before the collection thread must block is equal to (r * x) / 4. The syntax is in decimal units; the default is 5% and is formatted 0.05. Will be part of generated client configuration. io.sort.record.percent 0.05 io_sort_record_percent false
I/O Sort Spill Percent The soft limit in either the buffer or record collection buffers. When this limit is reached, a thread will begin to spill the contents to disk in the background. Note that this does not imply any chunking of data to the spill. A value less than 0.5 is not recommended. The syntax is in decimal units; the default is 80% and is formatted 0.8. Will be part of generated client configuration. io.sort.spill.percent 0.8 io_sort_spill_percent false
MapReduce Child Java Opts Base Java opts for the TaskTracker child processes. The following symbol, if present, will be interpolated: @taskid@ is replaced by current TaskID. Any other occurrences of '@' will go unchanged. For example, to enable verbose gc logging to a file named for the taskid in /tmp pass a value of: "-verbose:gc -Xloggc:/tmp/@taskid@.gc". The configuration variable 'mapred.child.ulimit' can be used to control the maximum virtual memory of the child processes. Note that unlike Hadoop, Cloudera Manager separates the child options into this setting and a separate setting just for the maximum heap size. Will be part of generated client configuration. mapred.child.java.opts -Djava.net.preferIPv4Stack=true mapred_child_java_opts_base false
Map Task Java Opts Base Java opts for the TaskTracker child map processes. The following symbol, if present, will be interpolated: @taskid@ is replaced by current TaskID. Any other occurrences of '@' will go unchanged. For example, to enable verbose gc logging to a file named for the taskid in /tmp pass a value of: "-verbose:gc -Xloggc:/tmp/@taskid@.gc". The configuration variable 'Map Task Maximum Virtual Memory' can be used to control the maximum virtual memory of the map processes. This takes precedence over the generic 'mapred.child.java.opts'. Will be part of generated client configuration. mapred.map.child.java.opts mapred_map_task_java_opts false
Default Number of Parallel Transfers During Shuffle The default number of parallel transfers run by reduce during the copy (shuffle) phase. This number should be between sqrt(nodes*number_of_map_slots_per_node) and nodes*number_of_map_slots_per_node/2. Will be part of generated client configuration. mapred.reduce.parallel.copies 10 mapred_reduce_parallel_copies false
Reduce Task Java Opts Base Java opts for the TaskTracker child reduce processes. The following symbol, if present, will be interpolated: @taskid@ is replaced by current TaskID. Any other occurrences of '@' will go unchanged. For example, to enable verbose gc logging to a file named for the taskid in /tmp pass a value of: "-verbose:gc -Xloggc:/tmp/@taskid@.gc". The configuration variable 'Reduce Task Maximum Virtual Memory' can be used to control the maximum virtual memory of the reduce processes. This takes precedence over the generic 'mapred.child.java.opts'. Will be part of generated client configuration. mapred.reduce.child.java.opts mapred_reduce_task_java_opts false

Resource Management

Display Name Description Related Name Default Value Unit API Name Required
MapReduce Child Java Maximum Heap Size The maximum heap size, in bytes, of the Java child process. This number will be formatted and concatenated with the 'base' setting for 'mapred.child.java.opts' to pass to Hadoop. Will be part of generated client configuration. 1 GiB mapred_child_java_opts_max_heap false
MapReduce Maximum Virtual Memory (KiB) The maximum virtual memory, in KiB, of a process launched by the MapReduce framework. This can be used to control both the MapReduce tasks and applications using Hadoop Pipes, Hadoop Streaming, and so on. By default, it is left unspecified to allow administrators to control it via 'limits.conf' and other mechanisms. Note: 'mapred.child.ulimit' must be greater than or equal to approximately 1.5 times the -Xmx passed to JavaVM, or else the VM might not start. Will be part of generated client configuration. mapred.child.ulimit mapred_child_ulimit false
Map Task Maximum Heap Size The maximum heap size, in bytes, of the child map processes. This number will be formatted and concatenated with 'Map Task Java Opts Base' to pass to Hadoop. Will be part of generated client configuration. mapred_map_task_max_heap false
Map Task Maximum Virtual Memory (KiB) The maximum virtual memory, in KiB, available to map tasks. Note: this must be greater than or equal to the -Xmx passed to the JavaVM via 'Map Task Java Opts', or else the VM might not start. This takes precedence over the generic 'mapred.child.ulimit'. Will be part of generated client configuration. mapred.map.child.ulimit mapred_map_task_ulimit false
Reduce Task Maximum Heap Size The maximum heap size, in bytes, of the child reduce processes. This number will be formatted and concatenated with 'Reduce Task Java Opts Base' to pass to Hadoop. Will be part of generated client configuration. mapred_reduce_task_max_heap false
Reduce Task Maximum Virtual Memory (KiB) The maximum virtual memory, in KiB, available to reduce tasks. Note: this must be greater than or equal to the -Xmx passed to the JavaVM via 'Map Task Java Opts', or else the VM might not start. This takes precedence over the generic 'mapred.child.ulimit'. Will be part of generated client configuration. mapred.reduce.child.ulimit mapred_reduce_task_ulimit false
Client Java Heap Size in Bytes Maximum size for the Java process heap memory. Passed to Java -Xmx. Measured in bytes. 256 MiB mapreduce_client_java_heapsize false

JobTracker Default Group

Advanced

Display Name Description Related Name Default Value Unit API Name Required
Hadoop Metrics Advanced Configuration Snippet (Safety Valve) Advanced Configuration Snippet (Safety Valve) for Hadoop Metrics. Properties will be inserted into hadoop-metrics.properties for this role only. Note that Cloudera Manager tunes hadoop-metrics.properties to work optimally with its Service Monitoring features. By overriding the default, Cloudera Manager might not be able to provide accurate monitoring information, health tests or alerts. hadoop_metrics_safety_valve false
JobTracker Advanced Configuration Snippet (Safety Valve) for mapred-site.xml For advanced use only, a string to be inserted into mapred-site.xml for this role only. jobtracker_config_safety_valve false
JobTracker Advanced Configuration Snippet (Safety Valve) for mapred_hosts_allow.txt For advanced use only, a string to be inserted into mapred_hosts_allow.txt for this role only. jobtracker_hosts_allow_safety_valve false
JobTracker Advanced Configuration Snippet (Safety Valve) for mapred_hosts_exclude.txt For advanced use only, a string to be inserted into mapred_hosts_exclude.txt for this role only. jobtracker_hosts_exclude_safety_valve false
Java Configuration Options for JobTracker These arguments will be passed as part of the Java command line. Commonly, garbage collection flags or extra debugging flags would be passed here. -XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:-CMSConcurrentMTEnabled -XX:CMSInitiatingOccupancyFraction=70 -XX:+CMSParallelRemarkEnabled jobtracker_java_opts false
JobTracker Logging Advanced Configuration Snippet (Safety Valve) For advanced use only, a string to be inserted into log4j.properties for this role only. log4j_safety_valve false
JobTracker Client Connection Retries The maximum number of times to retry between failovers. mapred.client.failover.connection.retries 0 mapred_client_failover_connection_retries false
JobTracker Client Max Retries The maximum number of times to retry on timeouts between failovers. mapred.client.failover.connection.retries.on.timeouts 0 mapred_client_failover_connection_retries_on_timeouts false
JobTracker Client Max Failover Attempt The maximum number of times a client of JobTracker tries to fail over. mapred.client.failover.max.attempts 15 mapred_client_failover_max_attempts false
JobTracker Client Base Sleep The time in milliseconds to wait before the first failover. mapred.client.failover.sleep.base.millis 500 millisecond(s) mapred_client_failover_sleep_base_millis false
JobTracker Client Maximum Sleep The maximum amount of time in milliseconds to wait between failovers (for exponential backoff). mapred.client.failover.sleep.max.millis 1.5 second(s) mapred_client_failover_sleep_max_millis false
Heap Dump Directory Path to directory where heap dumps are generated when java.lang.OutOfMemoryError error is thrown. This directory is automatically created if it doesn't exist. However, if this directory already exists, role user must have write access to this directory. If this directory is shared amongst multiple roles, it should have 1777 permissions. Note that the heap dump files are created with 600 permissions and are owned by the role user. The amount of free space in this directory should be greater than the maximum Java Process heap size configured for this role. /tmp oom_heap_dump_dir false
Dump Heap When Out of Memory When set, generates heap dump file when java.lang.OutOfMemoryError is thrown. false oom_heap_dump_enabled true
Kill When Out of Memory When set, a SIGKILL signal is sent to the role process when java.lang.OutOfMemoryError is thrown. true oom_sigkill_enabled true
Automatically Restart Process When set, this role's process is automatically (and transparently) restarted in the event of an unexpected failure. false process_auto_restart true

Classes

Display Name Description Related Name Default Value Unit API Name Required
Hadoop Socket Factory for Job Submission Socket Factory to use to connect to a MapReduce master (JobTracker). If null or empty, then use hadoop.rpc.socket.factory.class.default. hadoop.rpc.socket.factory.class.JobSubmissionProtocol hadoop_rpc_socket_factory_class_job_submission_protocol false
Task Scheduler The class responsible for scheduling tasks. Cloudera recommends the Fair Scheduler. The JobQueueTaskScheduler is often referred to as the FIFO scheduler. mapred.jobtracker.taskScheduler org.apache.hadoop.mapred.FairScheduler mapred_jobtracker_taskScheduler false

Jobs

Display Name Description Related Name Default Value Unit API Name Required
Capacity Scheduler Configuration Enter an XML string that represents the Capacity Scheduler configuration. <?xml version=1.0?> <!-- This is the configuration file for the resource manager in Hadoop. --> <!-- You can configure various scheduling parameters related to queues. --> <!-- The properties for a queue follow a naming convention, such as, --> <!-- mapred.capacity-scheduler.queue.<queue-name>.property-name. --> <configuration> <property> <name>mapred.capacity-scheduler.queue.default.capacity</name> <value>100</value> <description>Percentage of the number of slots in the cluster that are to be available for jobs in this queue. </description> </property> <property> <name>mapred.capacity-scheduler.queue.default.maximum-capacity</name> <value>-1</value> <description> maximum-capacity defines a limit beyond which a queue cannot use the capacity of the cluster. This provides a means to limit how much excess capacity a queue can use. By default, there is no limit. The maximum-capacity of a queue can only be greater than or equal to its minimum capacity. Default value of -1 implies a queue can use complete capacity of the cluster. This property could be to curtail certain jobs which are long running in nature from occupying more than a certain percentage of the cluster, which in the absence of pre-emption, could lead to capacity guarantees of other queues being affected. One important thing to note is that maximum-capacity is a percentage , so based on the cluster's capacity the max capacity would change. So if large no of nodes or racks get added to the cluster , max Capacity in absolute terms would increase accordingly. </description> </property> <property> <name>mapred.capacity-scheduler.queue.default.supports-priority</name> <value>false</value> <description>If true, priorities of jobs will be taken into account in scheduling decisions. </description> </property> <property> <name>mapred.capacity-scheduler.queue.default.minimum-user-limit-percent</name> <value>100</value> <description> Each queue enforces a limit on the percentage of resources allocated to a user at any given time, if there is competition for them. This user limit can vary between a minimum and maximum value. The former depends on the number of users who have submitted jobs, and the latter is set to this property value. For example, suppose the value of this property is 25. If two users have submitted jobs to a queue, no single user can use more than 50% of the queue resources. If a third user submits a job, no single user can use more than 33% of the queue resources. With 4 or more users, no user can use more than 25% of the queue's resources. A value of 100 implies no user limits are imposed. </description> </property> <property> <name>mapred.capacity-scheduler.queue.default.maximum-initialized-jobs-per-user</name> <value>2</value> <description>The maximum number of jobs to be pre-initialized for a user of the job queue. </description> </property> <!-- The default configuration settings for the capacity task scheduler --> <!-- The default values would be applied to all the queues which don't have --> <!-- the appropriate property for the particular queue --> <property> <name>mapred.capacity-scheduler.default-supports-priority</name> <value>false</value> <description>If true, priorities of jobs will be taken into account in scheduling decisions by default in a job queue. </description> </property> <property> <name>mapred.capacity-scheduler.default-minimum-user-limit-percent</name> <value>100</value> <description>The percentage of the resources limited to a particular user for the job queue at any given point of time by default. </description> </property> <property> <name>mapred.capacity-scheduler.default-maximum-initialized-jobs-per-user</name> <value>2</value> <description>The maximum number of jobs to be pre-initialized for a user of the job queue. </description> </property> <!-- Capacity scheduler Job Initialization configuration parameters --> <property> <name>mapred.capacity-scheduler.init-poll-interval</name> <value>5000</value> <description>The amount of time in miliseconds which is used to poll the job queues for jobs to initialize. </description> </property> <property> <name>mapred.capacity-scheduler.init-worker-threads</name> <value>5</value> <description>Number of worker threads which would be used by Initialization poller to initialize jobs in a set of queue. If number mentioned in property is equal to number of job queues then a single thread would initialize jobs in a queue. If lesser then a thread would get a set of queues assigned. If the number is greater then number of threads would be equal to number of job queues. </description> </property> </configuration> mapred_capacity_scheduler_configuration false
Fair Scheduler Allocation Enter an XML string that represents the Fair Scheduler allocation pools. <?xml version=1.0?> <allocations> </allocations> mapred_fairscheduler_allocation false
Fair Scheduler Allow Undeclared Pools Enable job submission to pools not declared in the allocation file. mapred.fairscheduler.allow.undeclared.pools true mapred_fairscheduler_allow_undeclared_pools false
Fair Scheduler Assign Multiple Tasks Allows the Fair Scheduler to assign both a map task and a reduce task on each Cloudera Agent heartbeat, which improves cluster throughput when there are many small tasks to run. mapred.fairscheduler.assignmultiple true mapred_fairscheduler_assignmultiple false
Fair Scheduler Pool Name Property Specify the 'jobconf' property that determines the pool that a job belongs in. The default is 'user.name' (one pool for each user). If you want to use MapReduce's "queue" system to enable authorization for the Fair Scheduler, specify 'mapred.job.queue.name'. This requires adding the Fair Scheduler's pool names to 'mapred.queue.names' and users to submit jobs using the 'mapred.job.queue.name' property instead of the 'mapred.fairscheduler.pool' property. Note that 'mapred.fairscheduler.poolnameproperty' is used only for jobs in which 'mapred.fairscheduler.pool' is not explicitly set. mapred.fairscheduler.poolnameproperty user.name mapred_fairscheduler_poolnameproperty false
Fair Scheduler Preemption Enables Fair Scheduler preemption. If a pool's minimum share is not met for some period of time, the Fair Scheduler optionally supports preemption of jobs in other pools. The pool will be allowed to kill tasks from other pools to make room to run. Preemption can be used to guarantee that production jobs are not starved while also allowing the Hadoop cluster to be used for experimental and research jobs. In addition, a pool can also be allowed to preempt tasks if it is below half of its fair share for a configurable timeout (generally set larger than the minimum share preemption timeout). When choosing tasks to kill, the Fair Scheduler picks the most-recently launched tasks from over-allocated jobs, to minimize wasted computation. Preemption does not cause the preempted jobs to fail because Hadoop jobs tolerate losing tasks; it only makes them take longer to finish. mapred.fairscheduler.preemption false mapred_fairscheduler_preemption false
Fair Scheduler Weight Adjuster An extension point that lets you specify a class to adjust the weights of running jobs. This class should implement the WeightAdjuster interface. There is currently one example implementation - NewJobWeightBooster, which increases the weight of jobs for their first 5 minutes to let short jobs finish faster. To use it, set the weightadjuster property to the full classname, org.apache.hadoop.mapred.NewJobWeightBooster. NewJobWeightBooster itself provides two parameters for setting the duration and boost factor. mapred.newjobweightbooster.factor: Factor by which new jobs weight should be boosted. Default is 3. mapred.newjobweightbooster.duration: Boost duration in milliseconds. Default is 300000 for 5 minutes. mapred.fairscheduler.weight.adjuster mapred_fairscheduler_weight_adjuster false
Persist JobTracker Job Status If enabled, job status information is persisted. mapred.job.tracker.persist.jobstatus.active false mapred_job_tracker_persist_jobstatus_active false
Directory for JobTracker Job Status Persistence The HDFS directory in which job status information is kept persistently. The directory must exist and be owned by the mapred user. mapred.job.tracker.persist.jobstatus.dir /jobtracker/jobsInfo mapred_job_tracker_persist_jobstatus_dir false
Time Limit of JobTracker Job Status Persistence The number of hours job status information is persisted in HDFS. The job status information will be available after it drops out of the memory queue and between JobTracker restarts. If zero is specified for this property, the job status information is not persisted. mapred.job.tracker.persist.jobstatus.hours 0 mapred_job_tracker_persist_jobstatus_hours false
Maximum Completed User Jobs The maximum number of completed jobs per user to retain before delegating them to the job history. mapred.jobtracker.completeuserjobs.maximum 5 mapred_jobtracker_completeuserjobs_maximum false
Enable Job Recovery Upon Restart Enables job recovery upon restart. If the property is set to true, then if and when the JobTracker stops while a job is running, it will resubmit the job on restart. mapred.jobtracker.restart.recover false mapred_jobtracker_restart_recover false
JobTracker Retire Job Interval (milliseconds) Number of milliseconds job history objects are kept. mapred.jobtracker.retirejob.interval 86400000 mapred_jobtracker_retirejob_interval false
MapReduce Queue Names Comma separated list of queues configured for the JobTracker in this service instance. Jobs are added to queues. Schedulers can configure different scheduling properties for the queues specified in this list. You can configure queue properties that are common to all schedulers, by using the naming convention 'mapred.queue.$QUEUE-NAME.$PROPERTY-NAME' in this property (for example, 'mapred.queue.default.submit-job-acl'). The number of queues configured in this property depends on the type of scheduler specified in 'mapred.jobtracker.taskScheduler'. The default scheduler JobQueueTaskScheduler supports a single queue only. Before adding more queues to this property, make sure that the scheduler in 'mapred.jobtracker.taskScheduler' supports multiple queues. This property can also be populated with the Fair Scheduler's pool names to enable authorization of the Fair Scheduler. This requires setting 'mapred.fairscheduler.poolnameproperty' to 'mapred.job.queue.name' and users to submit jobs to the right queue by setting the 'mapred.job.queue.name' property in their jobs. mapred.queue.names default mapred_queue_names_list false

Logs

Display Name Description Related Name Default Value Unit API Name Required
JobTracker Log Directory Directory where JobTracker will place its log files. hadoop.log.dir /var/log/hadoop-0.20-mapreduce jobtracker_log_dir false
JobTracker Logging Threshold The minimum log level for JobTracker logs INFO log_threshold false
JobTracker Maximum Log File Backups The maximum number of rolled log files to keep for JobTracker logs. Typically used by log4j. 10 max_log_backup_index false
JobTracker Max Log Size The maximum size, in megabytes, per log file for JobTracker logs. Typically used by log4j. 200 MiB max_log_size false

Metrics

Display Name Description Related Name Default Value Unit API Name Required
Hadoop Metrics Class Implementation daemons will use to report some internal statistics. The default (NoEmitMetricsContext) will display metrics on /metrics on the status port. The GangliaContext and GangliaContext31 classes will report metrics to your specified Ganglia Monitoring Daemons (gmond). The ganglia wire format changed incompatibly at version 3.1.0. If you are running any version of ganglia 3.1.0 or newer, use the GangliaContext31 metric class; otherwise, use the GangliaContext metric class. org.apache.hadoop.metrics.spi.NoEmitMetricsContext hadoop_metrics_class false
Hadoop Metrics Output Directory If using FileContext, directory to write metrics to. /tmp/metrics hadoop_metrics_dir false
Hadoop Metrics Ganglia Servers If using GangliaContext, a comma-delimited list of host:port pairs pointing to 'gmond' servers you would like to publish metrics to. In practice, this set of 'gmond' should match the set of 'gmond' in your 'gmetad' datasource list for the cluster. hadoop_metrics_ganglia_servers false

Monitoring

Display Name Description Related Name Default Value Unit API Name Required
Enable Health Alerts for this Role When set, Cloudera Manager will send alerts when the health of this role reaches the threshold specified by the EventServer setting eventserver_health_events_alert_threshold true enable_alerts false
Enable Configuration Change Alerts When set, Cloudera Manager will send alerts when this entity's configuration changes. false enable_config_alerts false
File Descriptor Monitoring Thresholds The health test thresholds of the number of file descriptors used. Specified as a percentage of file descriptor limit. Warning: 50.0 %, Critical: 70.0 % jobtracker_fd_thresholds false
Garbage Collection Duration Thresholds The health test thresholds for the weighted average time spent in Java garbage collection. Specified as a percentage of elapsed wall clock time. Warning: 30.0, Critical: 60.0 jobtracker_gc_duration_thresholds false
Garbage Collection Duration Monitoring Period The period to review when computing the moving average of garbage collection time. 5 minute(s) jobtracker_gc_duration_window false
JobTracker Host Health Test When computing the overall JobTracker health, consider the host's health. true jobtracker_host_health_enabled false
JobTracker Process Health Test Enables the health test that the JobTracker's process state is consistent with the role configuration true jobtracker_scm_health_enabled false
Health Check Startup Tolerance The amount of time allowed after this role is started that failures of health checks that rely on communication with this role will be tolerated. 5 minute(s) jobtracker_startup_tolerance false
Web Metric Collection Enables the health test that the Cloudera Manager Agent can successfully contact and gather metrics from the web server. true jobtracker_web_metric_collection_enabled false
Web Metric Collection Duration The health test thresholds on the duration of the metrics request to the web server. Warning: 10 second(s), Critical: Never jobtracker_web_metric_collection_thresholds false
Log Directory Free Space Monitoring Absolute Thresholds The health test thresholds for monitoring of free space on the filesystem that contains this role's log directory. Warning: 10 GiB, Critical: 5 GiB log_directory_free_space_absolute_thresholds false
Log Directory Free Space Monitoring Percentage Thresholds The health test thresholds for monitoring of free space on the filesystem that contains this role's log directory. Specified as a percentage of the capacity on that filesystem. This setting is not used if a Log Directory Free Space Monitoring Absolute Thresholds setting is configured. Warning: Never, Critical: Never log_directory_free_space_percentage_thresholds false
Rules to Extract Events from Log Files This file contains the rules which govern how log messages are turned into events by the custom log4j appender that this role loads. It is in JSON format, and is composed of a list of rules. Every log message is evaluated against each of these rules in turn to decide whether or not to send an event for that message. Each rule has some or all of the following fields:
  • alert - whether or not events generated from this rule should be promoted to alerts. A value of "true" will cause alerts to be generated. If not specified, the default is "false".
  • rate (mandatory) - the maximum number of log messages matching this rule that may be sent as events every minute. If more than rate matching log messages are received in a single minute, the extra messages are ignored. If rate is less than 0, the number of messages per minute is unlimited.
  • periodminutes - the number of minutes during which the publisher will only publish rate events or fewer. If not specified, the default is one minute
  • threshold - apply this rule only to messages with this log4j severity level or above. An example is "WARN" for warning level messages or higher.
  • content - match only those messages whose contents match this regular expression.
  • exceptiontype - match only those messages which are part of an exception message. The exception type must match this regular expression.
Example:{"alert": false, "rate": 10, "exceptiontype": "java.lang.StringIndexOutOfBoundsException"} This rule will send events to Cloudera Manager for every StringIndexOutOfBoundsException, up to a maximum of 10 every minute. 		version: 0, rules: [ alert: false, rate: 1, periodminutes: 1, threshold:FATAL, alert: false, rate: 0, threshold:WARN, content: .* is deprecated. Instead, use .*, alert: false, rate: 0, threshold:WARN, content: .* is deprecated. Use .* instead, alert: false, rate: 0, exceptiontype: java.io.IOException, alert: false, rate: 0, exceptiontype: java.net.SocketException, alert: false, rate: 0, exceptiontype: java.net.SocketClosedException, alert: false, rate: 0, exceptiontype: java.io.EOFException, alert: false, rate: 0, exceptiontype: java.nio.channels.CancelledKeyException, alert: false, rate: 1, periodminutes: 2, exceptiontype: .*, alert: false, rate: 0, threshold:WARN, content:Unknown job [^ ]+ being deleted.*, alert: false, rate: 0, threshold:WARN, content:Error executing shell command .+ No such process.+, alert: false, rate: 0, threshold:WARN, content:.*attempt to override final parameter.+, alert: false, rate: 0, threshold:WARN, content:[^ ]+ is a deprecated filesystem name. Use.*, alert: false, rate: 5, periodminutes: 1, threshold:WARN ] log_event_whitelist false
Role Triggers The configured triggers for this role. This is a JSON formatted list of triggers. These triggers are evaluated as part as the health system. Every trigger expression is parsed, and if the trigger condition is met, the list of actions provided in the trigger expression is executed. Each trigger has all of the following fields:
  • triggerName (mandatory) - the name of the trigger. This value must be unique for the specific role.
  • triggerExpression (mandatory) - a tsquery expression representing the trigger.
  • streamThreshold (optional) - the maximum number of streams that can satisfy a condition of a trigger before the condition fires. By default set to 0, and any stream returned will cause the condition to fire.
  • enabled (optional) - by default set to 'true'. If set to 'false' the trigger will not be evaluated.
For example, here is a JSON formatted trigger configured for a DataNode that fires if the DataNode has more than 1500 file-descriptors opened: 
[{"triggerName": "sample-trigger",
  "triggerExpression": "IF (SELECT fd_open WHERE roleName=$ROLENAME and last(fd_open) > 1500) DO health:red",
  "streamThreshold": 0}, "enabled": "true"]
Consult the trigger rules documentation for more details on how to write triggers using tsquery. The JSON format is evolving and may change in the future and as a result backward compatibility is not guaranteed between releases at this time. 		[] role_triggers true
Unexpected Exits Thresholds The health test thresholds for unexpected exits encountered within a recent period specified by the unexpected_exits_window configuration for the role. Warning: Never, Critical: Any unexpected_exits_thresholds false
Unexpected Exits Monitoring Period The period to review when computing unexpected exits. 5 minute(s) unexpected_exits_window false

Other

Display Name Description Related Name Default Value Unit API Name Required
JobTracker Logical Name For High Availability this is the logical name for the JobTracker active-standby pair. job_tracker_name false
JobTracker Local Data Directory Directory on the local filesystem where the JobTracker stores job configuration data. Directories that do not exist are ignored. A single directory is sufficient; a list of multiple directories will not cause problems. mapred.local.dir jobtracker_mapred_local_dir_list true
Job History Files Cleaner Interval Time interval for history cleaner to check for files to delete. Files are only deleted if they are older than mapreduce.jobhistory.max-age-ms. mapreduce.jobhistory.cleaner.interval 1 day(s) mapreduce_jobhistory_cleaner_interval false
Job History Files Maximum Age Job history files older than this time duration will deleted when the history cleaner runs. mapreduce.jobhistory.max-age-ms 7 day(s) mapreduce_jobhistory_max_age_ms false

Paths

Display Name Description Related Name Default Value Unit API Name Required
Running Job History Location Location to store the job history files of running jobs. This is a path on the host where the JobTracker is running. hadoop.job.history.location /var/log/hadoop-0.20-mapreduce/history hadoop_job_history_dir false
Completed Job History Location Location to store the job history files of completed jobs. If a location is not specified, the job history files of completed jobs are stored in a subdirectory of the 'Running Job History Location'. If set, completed jobs will be moved into this directory in HDFS. mapred.job.tracker.history.completed.location mapred_job_tracker_history_completed_dir false
MapReduce JobTracker Staging Root Directory The root HDFS directory of the staging area for users' MapReduce jobs; for example /user. The staging directories are always named after the user. mapreduce.jobtracker.staging.root.dir /user mapreduce_jobtracker_staging_root_dir false

Performance

Display Name Description Related Name Default Value Unit API Name Required
Hue Thrift Server Max Threadcount Maximum number of running threads for the Hue Thrift server running on the Jobtracker dfs.thrift.threads.max 20 dfs_thrift_threads_max false
Hue Thrift Server Min Threadcount Minimum number of running threads for the Hue Thrift server running on the Jobtracker dfs.thrift.threads.min 10 dfs_thrift_threads_min false
Hue Thrift Server Timeout Timeout in seconds for the Hue Thrift server running on the Jobtracker dfs.thrift.timeout 60 dfs_thrift_timeout false
JobTracker Handler Count The number of server threads for the JobTracker. This should be approximately 20 * ln(the number of TaskTracker nodes). mapred.job.tracker.handler.count 10 mapred_job_tracker_handler_count false
Maximum Tasks per Job The maximum number of tasks for a single job. Use a value of -1 B to specify no maximum. Note that allowing jobs with a large number of tasks increases memory usage by the JobTracker. mapred.jobtracker.maxtasks.per.job mapred_jobtracker_maxtasks_per_job false
User JobConf Limit The maximum allowed size of the user jobconf. mapred.user.jobconf.limit 5 MiB mapred_user_jobconf_limit false
JobTracker MetaInfo Maxsize The maximum permissible size of the split metainfo file. The JobTracker won't attempt to read split metainfo files bigger than the configured value. No limits if set to -1. mapreduce.jobtracker.split.metainfo.maxsize 10000000 mapreduce_jobtracker_split_metainfo_maxsize false
Maximum Process File Descriptors If configured, overrides the process soft and hard rlimits (also called ulimits) for file descriptors to the configured value. rlimit_fds false

Plugins

Display Name Description Related Name Default Value Unit API Name Required
Enable JobTracker Plugins Required for Hue If enabled, adds 'org.apache.hadoop.thriftfs.ThriftJobTrackerPlugin' to the 'mapred.jobtracker.plugins' configuration. This property must be enabled to allow Hue to operate. true hue_jobtracker_plugin false
MapReduce JobTracker Plugins mapred.jobtracker.plugins: Comma-separated list of JobTracker plugins to be activated. If one plugin cannot be loaded, all plugins are ignored. Note that there are separate controls below to enable the Hue Thrift plugin. mapred.jobtracker.plugins mapred_jobtracker_plugins_list false

Ports and Addresses

Display Name Description Related Name Default Value Unit API Name Required
JobTracker Port for HA Port of the High Availability service protocol for the JobTracker. The JobTracker listens on a separate port for High Availability operations which is why this property exists in addition to 'mapred.job.tracker'. mapred.ha.job.tracker 8023 ha_job_tracker_port false
Bind JobTracker to Wildcard Address If enabled, the JobTracker binds to the wildcard address ("0.0.0.0") on all of its ports. false job_tracker_bind_wildcard false
JobTracker Port Port for the internal JobTracker protocol. mapred.job.tracker 8021 job_tracker_port false
JobTracker HTTP Server Address The address where the JobTracker HTTP server listens. The default address, 0.0.0.0, binds to all interfaces. 0.0.0.0 mapred_job_tracker_http_host false
JobTracker HTTP Server Port The port where the JobTracker HTTP server listens. If the port is 0, the server starts on a free port. mapred.job.tracker.http.address 50030 mapred_job_tracker_http_port false
Hue Thrift Plugin Port Port to use for 'org.apache.hadoop.thriftfs.ThriftJobTrackerPlugin' that is used by Hue's NameNode plugin. jobtracker.thrift.address 9290 mapred_jobtracker_hue_thrift_plugin_port false

Resource Management

Display Name Description Related Name Default Value Unit API Name Required
Java Heap Size of Jobtracker in Bytes Maximum size for the Java Process heap memory. Passed to Java -Xmx. Measured in bytes. 1 GiB jobtracker_java_heapsize false
Cgroup CPU Shares Number of CPU shares to assign to this role. The greater the number of shares, the larger the share of the host's CPUs that will be given to this role when the host experiences CPU contention. Must be between 2 and 262144. Defaults to 1024 for processes not managed by Cloudera Manager. cpu.shares 1024 rm_cpu_shares true
Cgroup I/O Weight Weight for the read I/O requests issued by this role. The greater the weight, the higher the priority of the requests when the host experiences I/O contention. Must be between 100 and 1000. Defaults to 1000 for processes not managed by Cloudera Manager. blkio.weight 500 rm_io_weight true
Cgroup Memory Hard Limit Hard memory limit to assign to this role, enforced by the Linux kernel. When the limit is reached, the kernel will reclaim pages charged to the process. If reclaiming fails, the kernel may kill the process. Both anonymous as well as page cache pages contribute to the limit. Use a value of -1 B to specify no limit. By default processes not managed by Cloudera Manager will have no limit. memory.limit_in_bytes -1 MiB rm_memory_hard_limit true
Cgroup Memory Soft Limit Soft memory limit to assign to this role, enforced by the Linux kernel. When the limit is reached, the kernel will reclaim pages charged to the process if and only if the host is facing memory pressure. If reclaiming fails, the kernel may kill the process. Both anonymous as well as page cache pages contribute to the limit. Use a value of -1 B to specify no limit. By default processes not managed by Cloudera Manager will have no limit. memory.soft_limit_in_bytes -1 MiB rm_memory_soft_limit true

Security

Display Name Description Related Name Default Value Unit API Name Required
Enable MapReduce ACLs Specifies whether ACLs should be checked for authorization of users who are doing various queue and job-level operations. ACLs are disabled by default. If enabled, the JobTracker and TaskTracker perform access control checks when users make requests for queue and job operations. Examples of queue operations are submitting a job to the queue and killing a job in the queue. Examples of job operations are viewing the job details (mapreduce.job.acl-view-job), modifying the job (mapreduce.job.acl-modify-job), or using MapReduce APIs, RPCs, or the console and web user interfaces. mapred.acls.enabled false mapred_acls_enabled false
MapReduce Queue ACLs String representing an XML file that controls, per queue, which users are allowed to submit and administrate jobs in that queue. The default setting is that all users and groups are allowed to submit jobs to queue 'default' and no users or groups are allowed to administer jobs other than their own that are submitted to queue 'default'. <?xml version=1.0?> <?xml-stylesheet type=text/xsl href=configuration.xsl?> <configuration> <property> <name>mapred.queue.default.acl-submit-job</name> <value>*</value> </property> <property> <name>mapred.queue.default.acl-administer-jobs</name> <value> </value> </property> </configuration> mapred_queue_acls false
Web Interface Private Actions If enabled, administrative actions such as 'kill job' will be displayed in the JobTracker's web interface. These actions can then be triggered by anyone who has access to the web interface. webinterface.private.actions false webinterface_private_actions false

Service-Wide

Advanced

Display Name Description Related Name Default Value Unit API Name Required
System User's Home Directory The home directory of the system user on the local filesystem. This setting must reflect the system's configured value - only changing it here will not change the actual home directory. /var/lib/hadoop-mapreduce hdfs_user_home_dir true
MapReduce Service Advanced Configuration Snippet (Safety Valve) for hadoop-policy.xml For advanced use only, a string to be inserted into hadoop-policy.xml. Applies to configurations of all roles in this service except client configuration. mapreduce_hadoop_policy_config_safety_valve false
MapReduce Service Advanced Configuration Snippet (Safety Valve) for mapred-site.xml For advanced use only, a string to be inserted into mapred-site.xml. Applies to configurations of all roles in this service except client configuration. mapreduce_service_config_safety_valve false
MapReduce Service Environment Advanced Configuration Snippet (Safety Valve) For advanced use only, key-value pairs (one on each line) to be inserted into a role's environment. Applies to configurations of all roles in this service except client configuration. mapreduce_service_env_safety_valve false
System Group The group that this service's processes should run as. hadoop process_groupname true
System User The user that this service's processes should run as. mapred process_username true

Monitoring

Display Name Description Related Name Default Value Unit API Name Required
Enable Log Event Capture When set, each role will identify important log events and forward them to Cloudera Manager. true catch_events false
Enable Service Level Health Alerts When set, Cloudera Manager will send alerts when the health of this service reaches the threshold specified by the EventServer setting eventserver_health_events_alert_threshold true enable_alerts false
Enable Configuration Change Alerts When set, Cloudera Manager will send alerts when this entity's configuration changes. false enable_config_alerts false
Failover Controllers Healthy Enables the health check that verifies that the failover controllers associated with this service are healthy and running. true failover_controllers_healthy_enabled false
Activity Duration Rules To generate an event when certain activities are running slowly, enter rules for the activities in this setting. The syntax for a rule is '<regex>=<number>' where <number> is in minutes. Enter one rule per line in this text box. When a new activity starts, each <regex> expression is tested against the name of the activity for a match. The first rule that matches is used. If an activity matches a rule and runs longer than the <number> of minutes, an event will be sent. firehose_activity_duration_rules false
Alert on Activity Failure If enabled, an alert will be generated when any activity fails. true firehose_activity_failure_alert false
Alert on Slow Activities If enabled, an alert will be generated when an activity has been running longer than the duration specified in the 'Activity Duration Rules' setting. true firehose_activity_slow_alert false
Log Event Retry Frequency The frequency in which the log4j event publication appender will retry sending undelivered log events to the Event server, in seconds 30 log_event_retry_frequency false
Active JobTracker Detection Window The tolerance window that will be used in Mapreduce service tests that depend on detection of the active JobTracker. 3 minute(s) mapreduce_active_jobtracker_detecton_window false
JobTracker Activation Startup Tolerance The amount of time after JobTracker(s) start that the lack of an active JobTracker will be tolerated. This is intended to allow either the auto-failover daemon to make a JobTracker active, or a specifically issued failover command to take effect. This is an advanced option that does not often need to be changed. 3 minute(s) mapreduce_jobtracker_activation_startup_tolerance false
JobTracker Role Health Test When computing the overall MapReduce cluster health, consider the JobTracker's health true mapreduce_jobtracker_health_enabled false
Standby JobTracker Health Test When computing the overall cluster health, consider the health of the standby JobTracker. true mapreduce_standby_jobtrackers_health_enabled false
Healthy TaskTracker Monitoring Thresholds The health test thresholds of the overall TaskTracker health. The check returns "Concerning" health if the percentage of "Healthy" TaskTrackers falls below the warning threshold. The check is unhealthy if the total percentage of "Healthy" and "Concerning" TaskTrackers falls below the critical threshold. Warning: 95.0 %, Critical: 90.0 % mapreduce_tasktrackers_healthy_thresholds false
Service Triggers The configured triggers for this service. This is a JSON formatted list of triggers. These triggers are evaluated as part as the health system. Every trigger expression is parsed, and if the trigger condition is met, the list of actions provided in the trigger expression is executed. Each trigger has all of the following fields:
  • triggerName (mandatory) - the name of the trigger. This value must be unique for the specific service.
  • triggerExpression (mandatory) - a tsquery expression representing the trigger.
  • streamThreshold (optional) - the maximum number of streams that can satisfy a condition of a trigger before the condition fires. By default set to 0, and any stream returned will cause the condition to fire.
  • enabled (optional) - by default set to 'true'. If set to 'false' the trigger will not be evaluated.
For example, here is a JSON formatted trigger that fires if there are more than 10 DataNodes with more than 500 file-descriptors opened: 
[{"triggerName": "sample-trigger",
  "triggerExpression": "IF (SELECT fd_open WHERE roleType = DataNode and last(fd_open) > 500) DO health:red",
  "streamThreshold": 10, "enabled": "true"}]
Consult the trigger rules documentation for more details on how to write triggers using tsquery. The JSON format is evolving and may change in the future and as a result backward compatibility is not guaranteed between releases at this time. 		[] service_triggers true
Service Monitor Client Config Overrides For advanced use only, a list of configuration properties that will be used by the Service Monitor instead of the current client configuration for the service. <property><name>mapreduce.jobclient.rpc.timeout</name><value>10000</value></property><property><name>ipc.ping.interval</name><value>10000</value></property><property><name>ipc.client.connect.timeout</name><value>10000</value></property><property><name>ipc.client.connect.max.retries</name><value>0</value></property><property><name>ipc.client.connect.max.retries.on.timeouts</name><value>0</value></property><property><name>mapreduce.job.counters.limit</name><value>12000</value></property><property><name>mapreduce.job.counters.max</name><value>12000</value></property><property><name>mapreduce.job.counters.group.name.max</name><value>12800</value></property><property><name>mapreduce.job.counters.counter.name.max</name><value>12800</value></property><property><name>mapreduce.job.counters.groups.max</name><value>5000</value></property> smon_client_config_overrides false
Service Monitor Derived Configs Advanced Configuration Snippet (Safety Valve) For advanced use only, a list of derived configuration properties that will be used by the Service Monitor instead of the default ones. smon_derived_configs_safety_valve false

Other

Display Name Description Related Name Default Value Unit API Name Required
HDFS Service Name of the HDFS service that this MapReduce service instance depends on hdfs_service true
ZooKeeper Service Name of the ZooKeeper service that this MapReduce service instance depends on zookeeper_service false

Paths

Display Name Description Related Name Default Value Unit API Name Required
MapReduce System Directory The HDFS directory where the MapReduce service stores system files. This directory must be accessible from both the server and client machines. For example: /hadoop/mapred/system/ mapred.system.dir /tmp/mapred/system mapred_system_dir false

Performance

Display Name Description Related Name Default Value Unit API Name Required
Enable HDFS Short Circuit Read Enable HDFS short circuit read. This allows a client co-located with the DataNode to read HDFS file blocks directly. This gives a performance boost to distributed clients that are aware of locality. dfs.client.read.shortcircuit false dfs_client_read_shortcircuit false
SequenceFile I/O Buffer Size Size of buffer for read and write operations of SequenceFiles. io.file.buffer.size 64 KiB io_file_buffer_size false
Job Counters Limit Limit on the number of counters allowed per job. mapreduce.job.counters.max 120 mapreduce_job_counters_limit false

Proxy

Display Name Description Related Name Default Value Unit API Name Required
HDFS Proxy User Groups Comma-delimited list of groups that you want to allow the HDFS user to impersonate. The default '*' allows all groups. To disable entirely, use a string that doesn't correspond to a group name, such as '_no_group_'. hadoop.proxyuser.hdfs.groups * hdfs_proxy_user_groups_list false
HDFS Proxy User Hosts Comma-delimited list of hosts where you want to allow the HDFS user to impersonate other users. The default '*' allows all hosts. To disable entirely, use a string that doesn't correspond to a host name, such as '_no_host'. hadoop.proxyuser.hdfs.hosts * hdfs_proxy_user_hosts_list false
Hive Proxy User Groups Comma-delimited list of groups that you want to allow the Hive user to impersonate. The default '*' allows all groups. To disable entirely, use a string that doesn't correspond to a group name, such as '_no_group_'. hadoop.proxyuser.hive.groups * hive_proxy_user_groups_list false
Hive Proxy User Hosts Comma-delimited list of hosts where you want to allow the Hive user to impersonate other users. The default '*' allows all hosts. To disable entirely, use a string that doesn't correspond to a host name, such as '_no_host'. hadoop.proxyuser.hive.hosts * hive_proxy_user_hosts_list false
Hue Proxy User Groups Comma-delimited list of groups that you want to allow the Hue user to impersonate. The default '*' allows all groups. To disable entirely, use a string that doesn't correspond to a group name, such as '_no_group_'. hadoop.proxyuser.hue.groups * hue_proxy_user_groups_list false
Hue Proxy User Hosts Comma-delimited list of hosts where you want to allow the Hue user to impersonate other users. The default '*' allows all hosts. To disable entirely, use a string that doesn't correspond to a host name, such as '_no_host'. hadoop.proxyuser.hue.hosts * hue_proxy_user_hosts_list false
Oozie Proxy User Groups Allows the oozie superuser to impersonate any members of a comma-delimited list of groups. The default '*' allows all groups. To disable entirely, use a string that doesn't correspond to a group name, such as '_no_group_'. hadoop.proxyuser.oozie.groups * oozie_proxy_user_groups_list false
Oozie Proxy User Hosts Comma-delimited list of hosts where you want to allow the oozie user to impersonate other users. The default '*' allows all hosts. To disable entirely, use a string that doesn't correspond to a host name, such as '_no_host'. hadoop.proxyuser.oozie.hosts * oozie_proxy_user_hosts_list false

Security

Display Name Description Related Name Default Value Unit API Name Required
Enable Authentication for HTTP Web-Consoles Enables authentication for hadoop HTTP web-consoles for all roles of this service. Note: This is effective only if security is enabled for the HDFS service. false hadoop_secure_web_ui false
Hue's Kerberos Principal Short Name The short name of Hue's Kerberos principal hue.kerberos.principal.shortname hue hue_kerberos_principal_shortname false

TaskTracker Default Group

Advanced

Display Name Description Related Name Default Value Unit API Name Required
Hadoop Metrics Advanced Configuration Snippet (Safety Valve) Advanced Configuration Snippet (Safety Valve) for Hadoop Metrics. Properties will be inserted into hadoop-metrics.properties for this role only. Note that Cloudera Manager tunes hadoop-metrics.properties to work optimally with its Service Monitoring features. By overriding the default, Cloudera Manager might not be able to provide accurate monitoring information, health tests or alerts. hadoop_metrics_safety_valve false
TaskTracker Logging Advanced Configuration Snippet (Safety Valve) For advanced use only, a string to be inserted into log4j.properties for this role only. log4j_safety_valve false
Healthchecker Script Arguments Comma-separated list of arguments which are to be passed to node health script when it is being launched. mapred.healthChecker.script.args mapred_healthchecker_script_args false
Healthchecker Script Path Absolute path to the script which is periodically run by the node health monitoring service to determine if the node is healthy or not. If the value of this key is empty or the file does not exist in the location configured here, the node health monitoring service is not started. mapred.healthChecker.script.path mapred_healthchecker_script_path false
Heap Dump Directory Path to directory where heap dumps are generated when java.lang.OutOfMemoryError error is thrown. This directory is automatically created if it doesn't exist. However, if this directory already exists, role user must have write access to this directory. If this directory is shared amongst multiple roles, it should have 1777 permissions. Note that the heap dump files are created with 600 permissions and are owned by the role user. The amount of free space in this directory should be greater than the maximum Java Process heap size configured for this role. /tmp oom_heap_dump_dir false
Dump Heap When Out of Memory When set, generates heap dump file when java.lang.OutOfMemoryError is thrown. false oom_heap_dump_enabled true
Kill When Out of Memory When set, a SIGKILL signal is sent to the role process when java.lang.OutOfMemoryError is thrown. true oom_sigkill_enabled true
Automatically Restart Process When set, this role's process is automatically (and transparently) restarted in the event of an unexpected failure. true process_auto_restart true
TaskTracker Advanced Configuration Snippet (Safety Valve) for taskcontroller.cfg For advanced use only, a string to be inserted into taskcontroller.cfg for this role only. taskcontroller_config_safety_valve false
TaskTracker Advanced Configuration Snippet (Safety Valve) for mapred-site.xml For advanced use only, a string to be inserted into mapred-site.xml for this role only. tasktracker_config_safety_valve false
Java Configuration Options for TaskTracker These arguments will be passed as part of the Java command line. Commonly, garbage collection flags or extra debugging flags would be passed here. -XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:-CMSConcurrentMTEnabled -XX:CMSInitiatingOccupancyFraction=70 -XX:+CMSParallelRemarkEnabled tasktracker_java_opts false

Classes

Display Name Description Related Name Default Value Unit API Name Required
TaskTracker Instrumentation Class The instrumentation class to associate with each TaskTracker. If using Cloudera's Activity Monitor, adjust this to use org.apache.hadoop.mapred.TaskTrackerCmonInst. mapred.tasktracker.instrumentation org.apache.hadoop.mapred.TaskTrackerMetricsInst mapred_tasktracker_instrumentation false

Compression

Display Name Description Related Name Default Value Unit API Name Required
Compression Codecs (Client Override) Comma-separated list of compression codecs that can be used in job or map compression. io.compression.codecs override_io_compression_codecs false
Use Compression on Map Outputs (Client Override) If enabled, uses compression on the map outputs before they are sent across the network. Will override value in client configuration. mapred.compress.map.output no_override override_mapred_compress_map_output false
Compression Codec of MapReduce Map Output (Client Override) For MapReduce map outputs that are compressed, specify the compression codec to use. Will override value in client configuration. mapred.map.output.compression.codec override_mapred_map_output_compression_codec false
Compress MapReduce Job Output (Client Override) Compress the output of MapReduce jobs. Will override value in client configuration. mapred.output.compress no_override override_mapred_output_compress false
Compression Codec of MapReduce Job Output (Client Override) For MapReduce job outputs that are compressed, specify the compression codec to use. Will override value in client configuration. mapred.output.compression.codec override_mapred_output_compression_codec false
Compression Type of MapReduce Job Output (Client Override) For MapReduce job outputs that are compressed as SequenceFiles, you can select one of these compression type options: NONE, RECORD or BLOCK. Cloudera recommends BLOCK. Will override value in client configuration. mapred.output.compression.type override_mapred_output_compression_type false

Jobs

Display Name Description Related Name Default Value Unit API Name Required
Number of Tasks to Run per JVM (Client Override) Number of tasks to run per JVM. If set to -1, there is no limit. Will override value in client configuration. mapred.job.reuse.jvm.num.tasks override_mapred_job_reuse_jvm_num_tasks false
Map Tasks Speculative Execution (Client Override) If enabled, multiple instances of some map tasks may be executed in parallel. mapred.map.tasks.speculative.execution no_override override_mapred_map_tasks_speculative_execution false
Number of Map Tasks to Complete Before Reduce Tasks (Client Override) Fraction of the number of map tasks in the job which should be completed before reduce tasks are scheduled for the job. mapred.reduce.slowstart.completed.maps override_mapred_reduce_slowstart_completed_maps false
Reduce Tasks Speculative Execution (Client Override) If enabled, multiple instances of some reduce tasks may be executed in parallel. mapred.reduce.tasks.speculative.execution no_override override_mapred_reduce_tasks_speculative_execution false
Mapreduce Submit Replication (Client Override) The replication level for submitted job files. mapred.submit.replication override_mapred_submit_replication false
Maximum Time to Retain User Logs (Client Override) The maximum time, in hours, to retain the user logs after job completion. mapred.userlog.retain.hours override_mapred_userlog_retain_hours false

Logs

Display Name Description Related Name Default Value Unit API Name Required
TaskTracker Logging Threshold The minimum log level for TaskTracker logs INFO log_threshold false
TaskTracker Maximum Log File Backups The maximum number of rolled log files to keep for TaskTracker logs. Typically used by log4j. 10 max_log_backup_index false
TaskTracker Max Log Size The maximum size, in megabytes, per log file for TaskTracker logs. Typically used by log4j. 200 MiB max_log_size false
TaskTracker Log Directory Directory where TaskTracker will place its log files. hadoop.log.dir /var/log/hadoop-0.20-mapreduce tasktracker_log_dir false

Metrics

Display Name Description Related Name Default Value Unit API Name Required
Hadoop Metrics Class Implementation daemons will use to report some internal statistics. The default (NoEmitMetricsContext) will display metrics on /metrics on the status port. The GangliaContext and GangliaContext31 classes will report metrics to your specified Ganglia Monitoring Daemons (gmond). The ganglia wire format changed incompatibly at version 3.1.0. If you are running any version of ganglia 3.1.0 or newer, use the GangliaContext31 metric class; otherwise, use the GangliaContext metric class. org.apache.hadoop.metrics.spi.NoEmitMetricsContext hadoop_metrics_class false
Hadoop Metrics Output Directory If using FileContext, directory to write metrics to. /tmp/metrics hadoop_metrics_dir false
Hadoop Metrics Ganglia Servers If using GangliaContext, a comma-delimited list of host:port pairs pointing to 'gmond' servers you would like to publish metrics to. In practice, this set of 'gmond' should match the set of 'gmond' in your 'gmetad' datasource list for the cluster. hadoop_metrics_ganglia_servers false

Monitoring

Display Name Description Related Name Default Value Unit API Name Required
Enable Health Alerts for this Role When set, Cloudera Manager will send alerts when the health of this role reaches the threshold specified by the EventServer setting eventserver_health_events_alert_threshold false enable_alerts false
Enable Configuration Change Alerts When set, Cloudera Manager will send alerts when this entity's configuration changes. false enable_config_alerts false
Log Directory Free Space Monitoring Absolute Thresholds The health test thresholds for monitoring of free space on the filesystem that contains this role's log directory. Warning: 10 GiB, Critical: 5 GiB log_directory_free_space_absolute_thresholds false
Log Directory Free Space Monitoring Percentage Thresholds The health test thresholds for monitoring of free space on the filesystem that contains this role's log directory. Specified as a percentage of the capacity on that filesystem. This setting is not used if a Log Directory Free Space Monitoring Absolute Thresholds setting is configured. Warning: Never, Critical: Never log_directory_free_space_percentage_thresholds false
Rules to Extract Events from Log Files This file contains the rules which govern how log messages are turned into events by the custom log4j appender that this role loads. It is in JSON format, and is composed of a list of rules. Every log message is evaluated against each of these rules in turn to decide whether or not to send an event for that message. Each rule has some or all of the following fields:
  • alert - whether or not events generated from this rule should be promoted to alerts. A value of "true" will cause alerts to be generated. If not specified, the default is "false".
  • rate (mandatory) - the maximum number of log messages matching this rule that may be sent as events every minute. If more than rate matching log messages are received in a single minute, the extra messages are ignored. If rate is less than 0, the number of messages per minute is unlimited.
  • periodminutes - the number of minutes during which the publisher will only publish rate events or fewer. If not specified, the default is one minute
  • threshold - apply this rule only to messages with this log4j severity level or above. An example is "WARN" for warning level messages or higher.
  • content - match only those messages whose contents match this regular expression.
  • exceptiontype - match only those messages which are part of an exception message. The exception type must match this regular expression.
Example:{"alert": false, "rate": 10, "exceptiontype": "java.lang.StringIndexOutOfBoundsException"} This rule will send events to Cloudera Manager for every StringIndexOutOfBoundsException, up to a maximum of 10 every minute. 		version: 0, rules: [ alert: false, rate: 0, threshold:ERROR, content:/mapOutput.*, alert: false, rate: 1, periodminutes: 1, threshold:FATAL, alert: false, rate: 0, threshold:WARN, content: .* is deprecated. Instead, use .*, alert: false, rate: 0, threshold:WARN, content: .* is deprecated. Use .* instead, alert: false, rate: 0, exceptiontype: java.io.IOException, alert: false, rate: 0, exceptiontype: java.net.SocketException, alert: false, rate: 0, exceptiontype: java.net.SocketClosedException, alert: false, rate: 0, exceptiontype: java.io.EOFException, alert: false, rate: 0, exceptiontype: org.mortbay.jetty.EofException, alert: false, rate: 0, exceptiontype: java.nio.channels.CancelledKeyException, alert: false, rate: 1, periodminutes: 2, exceptiontype: .*, alert: false, rate: 0, threshold:WARN, content:Unknown job [^ ]+ being deleted.*, alert: false, rate: 0, threshold:WARN, content:Error executing shell command .+ No such process.+, alert: false, rate: 0, threshold:WARN, content:Error sending signal TERM to process group.*No such process.*, alert: false, rate: 0, threshold:WARN, content:Exit code from task is :.*, alert: false, rate: 0, threshold:WARN, content:TaskTracker's totalMemoryAllottedForTasks is -1. TaskMemoryManager is disabled., alert: false, rate: 0, threshold:WARN, content:.*attempt to override final parameter.+, alert: false, rate: 0, threshold:WARN, content:[^ ]+ is a deprecated filesystem name. Use.*, alert: false, rate: 1, threshold:INFO, content:.*failed to report status for.*Killing!, alert: false, rate: 1, periodminutes: 1, threshold:WARN ] log_event_whitelist false
Role Triggers The configured triggers for this role. This is a JSON formatted list of triggers. These triggers are evaluated as part as the health system. Every trigger expression is parsed, and if the trigger condition is met, the list of actions provided in the trigger expression is executed. Each trigger has all of the following fields:
  • triggerName (mandatory) - the name of the trigger. This value must be unique for the specific role.
  • triggerExpression (mandatory) - a tsquery expression representing the trigger.
  • streamThreshold (optional) - the maximum number of streams that can satisfy a condition of a trigger before the condition fires. By default set to 0, and any stream returned will cause the condition to fire.
  • enabled (optional) - by default set to 'true'. If set to 'false' the trigger will not be evaluated.
For example, here is a JSON formatted trigger configured for a DataNode that fires if the DataNode has more than 1500 file-descriptors opened: 
[{"triggerName": "sample-trigger",
  "triggerExpression": "IF (SELECT fd_open WHERE roleName=$ROLENAME and last(fd_open) > 1500) DO health:red",
  "streamThreshold": 0}, "enabled": "true"]
Consult the trigger rules documentation for more details on how to write triggers using tsquery. The JSON format is evolving and may change in the future and as a result backward compatibility is not guaranteed between releases at this time. 		[] role_triggers true
TaskTracker Blacklisted Health Test Enables the health test that the TaskTracker is not blacklisted true tasktracker_blacklisted_health_enabled false
TaskTracker Connectivity Health Test Enables the health test that the TaskTracker is connected to the JobTracker true tasktracker_connectivity_health_enabled false
TaskTracker Connectivity Tolerance at Startup The amount of time to wait for the TaskTracker to fully start up and connect to the JobTracker before enforcing the connectivity check. 3 minute(s) tasktracker_connectivity_tolerance false
File Descriptor Monitoring Thresholds The health test thresholds of the number of file descriptors used. Specified as a percentage of file descriptor limit. Warning: 50.0 %, Critical: 70.0 % tasktracker_fd_thresholds false
Garbage Collection Duration Thresholds The health test thresholds for the weighted average time spent in Java garbage collection. Specified as a percentage of elapsed wall clock time. Warning: 30.0, Critical: 60.0 tasktracker_gc_duration_thresholds false
Garbage Collection Duration Monitoring Period The period to review when computing the moving average of garbage collection time. 5 minute(s) tasktracker_gc_duration_window false
TaskTracker Host Health Test When computing the overall TaskTracker health, consider the host's health. true tasktracker_host_health_enabled false
TaskTracker Process Health Test Enables the health test that the TaskTracker's process state is consistent with the role configuration true tasktracker_scm_health_enabled false
Web Metric Collection Enables the health test that the Cloudera Manager Agent can successfully contact and gather metrics from the web server. true tasktracker_web_metric_collection_enabled false
Web Metric Collection Duration The health test thresholds on the duration of the metrics request to the web server. Warning: 10 second(s), Critical: Never tasktracker_web_metric_collection_thresholds false
Unexpected Exits Thresholds The health test thresholds for unexpected exits encountered within a recent period specified by the unexpected_exits_window configuration for the role. Warning: Never, Critical: Any unexpected_exits_thresholds false
Unexpected Exits Monitoring Period The period to review when computing unexpected exits. 5 minute(s) unexpected_exits_window false

Other

Display Name Description Related Name Default Value Unit API Name Required
TaskTracker Local Data Directory List List of directories on the local filesystem where a TaskTracker stores intermediate data files. To spread disk I/O, enter a comma-separated list of directories on different devices. Directories that do not exist are ignored. Typical values are /data/N/mapred/local for N = 1, 2, 3... mapred.local.dir tasktracker_mapred_local_dir_list true

Performance

Display Name Description Related Name Default Value Unit API Name Required
I/O Sort Factor (Client Override) The number of streams to merge at once while sorting files. That is, the number of sort heads to use during the merge sort on the reducer side. This determines the number of open file handles. Merging more files in parallel reduces merge sort iterations and improves run time by eliminating disk I/O. Note that merging more files in parallel uses more memory. If 'io.sort.factor' is set too high or the maximum JVM heap is set too low, excessive garbage collection will occur. The Hadoop default is 10, but Cloudera recommends a higher value. Will override value in client configuration. io.sort.factor override_io_sort_factor false
I/O Sort Memory Buffer (MiB) (Client Override) The total amount of memory buffer, in megabytes, to use while sorting files. Note that this memory comes out of the user JVM heap size (meaning total user JVM heap - this amount of memory = total user usable heap space. Note that Cloudera's default differs from Hadoop's default; Cloudera uses a bigger buffer by default because modern machines often have more RAM. Will override value in client configuration. io.sort.mb override_io_sort_mb false
I/O Sort Record Percent (Client Override) The percentage of 'io.sort.mb' dedicated to tracking record boundaries. If this value is represented as 'r', and 'io.sort.mb' is represented as 'x', then the maximum number of records collected before the collection thread must block is equal to (r * x) / 4. The syntax is in decimal units; the default is 5% and is formatted 0.05. Will override value in client configuration. io.sort.record.percent override_io_sort_record_percent false
I/O Sort Spill Percent (Client Override) The soft limit in either the buffer or record collection buffers. When this limit is reached, a thread will begin to spill the contents to disk in the background. Note that this does not imply any chunking of data to the spill. A value less than 0.5 is not recommended. The syntax is in decimal units; the default is 80% and is formatted 0.8. Will override value in client configuration. io.sort.spill.percent override_io_sort_spill_percent false
MapReduce Child Java Opts Base (Client Override) Java opts for the TaskTracker child processes. The following symbol, if present, will be interpolated: @taskid@ is replaced by current TaskID. Any other occurrences of '@' will go unchanged. For example, to enable verbose gc logging to a file named for the taskid in /tmp pass a value of: "-verbose:gc -Xloggc:/tmp/@taskid@.gc". The configuration variable 'mapred.child.ulimit' can be used to control the maximum virtual memory of the child processes. Note that unlike Hadoop, Cloudera Manager separates the child options into this setting and a separate setting just for the maximum heap size. Will override value in client configuration. mapred.child.java.opts override_mapred_child_java_opts_base false
Map Task Java Opts Base (Client Override) Java opts for the TaskTracker child map processes. The following symbol, if present, will be interpolated: @taskid@ is replaced by current TaskID. Any other occurrences of '@' will go unchanged. For example, to enable verbose gc logging to a file named for the taskid in /tmp pass a value of: "-verbose:gc -Xloggc:/tmp/@taskid@.gc". The configuration variable 'Map Task Maximum Virtual Memory' can be used to control the maximum virtual memory of the map processes. This takes precedence over the generic 'mapred.child.java.opts'. mapred.map.child.java.opts override_mapred_map_task_java_opts false
Default Number of Parallel Transfers During Shuffle (Client Override) The default number of parallel transfers run by reduce during the copy (shuffle) phase. This number should be between sqrt(nodes*number_of_map_slots_per_node) and nodes*s/2. Will override value in client configuration. mapred.reduce.parallel.copies override_mapred_reduce_parallel_copies false
Reduce Task Java Opts Base (Client Override) Java opts for the TaskTracker child map processes. The following symbol, if present, will be interpolated: @taskid@ is replaced by current TaskID. Any other occurrences of '@' will go unchanged. For example, to enable verbose gc logging to a file named for the taskid in /tmp pass a value of: "-verbose:gc -Xloggc:/tmp/@taskid@.gc". The configuration variable 'Reduce Task Maximum Virtual Memory' can be used to control the maximum virtual memory of the reduce processes. This takes precedence over the generic 'mapred.child.java.opts'. mapred.reduce.child.java.opts override_mapred_reduce_task_java_opts false
Maximum Process File Descriptors If configured, overrides the process soft and hard rlimits (also called ulimits) for file descriptors to the configured value. rlimit_fds false
Number of TaskTracker HTTP Threads The number of worker threads for the HTTP server. This is used for map output fetching. tasktracker.http.threads 80 tasktracker_http_threads false

Ports and Addresses

Display Name Description Related Name Default Value Unit API Name Required
TaskTracker Activity Monitor Instrumentation Plugin Address Address where TaskTracker Activity Monitor instrumentation plugin listens for requests. This setting is ignored unless the TaskTracker Instrumentation Class is set to org.apache.hadoop.mapred.TaskTrackerCmonInst. This is usually set to 127.0.0.1. mapred.tasktracker.instrumentation.cmon.jettyhost 127.0.0.1 mapred_tasktracker_instrumentation_cmon_jettyhost false
TaskTracker Activity Monitor Instrumentation Plugin Port Port where TaskTracker Activity Monitor instrumentation plugin listens for requests. This setting is ignored unless the TaskTracker Instrumentation Class Class is set to org.apache.hadoop.mapred.TaskTrackerCmonInst. mapred.tasktracker.instrumentation.cmon.jettyport 4867 mapred_tasktracker_instrumentation_cmon_jettyport false
TaskTracker Web UI Address Address where TaskTracker listens for web requests 0.0.0.0 task_tracker_http_address false
TaskTracker Web UI Port Port where TaskTracker listens for web requests mapred.task.tracker.http.address 50060 task_tracker_http_port false

Resource Management

Display Name Description Related Name Default Value Unit API Name Required
Maximum Number of Simultaneous Map Tasks The maximum number of map tasks that a TaskTracker can run simultaneously. Sometimes referred to as "map slots." mapred.tasktracker.map.tasks.maximum 2 mapred_tasktracker_map_tasks_maximum false
Maximum Number of Simultaneous Reduce Tasks The maximum number of reduce tasks that a TaskTracker can run simultaneously. Sometimes referred to as "reduce slots." mapred.tasktracker.reduce.tasks.maximum 2 mapred_tasktracker_reduce_tasks_maximum false
MapReduce Child Java Maximum Heap Size (Client Override) The maximum heap size, in bytes, of the Java child process. This number will be formatted and concatenated with the 'base' setting for 'mapred.child.java.opts' to pass to Hadoop. Will override value in client configuration. override_mapred_child_java_opts_max_heap false
MapReduce Maximum Virtual Memory (KiB) (Client Override) The maximum virtual memory, in KiB, of a process launched by the MapReduce framework. This can be used to control both the MapReduce tasks and applications using Hadoop Pipes, Hadoop Streaming, and so on. By default, it is left unspecified to allow administrators to control it 'via limits.conf' and other mechanisms. Note: 'mapred.child.ulimit' must be greater than or equal to approximately 1.5 times the -Xmx passed to JavaVM, or else the VM might not start. Will override value in client configuration. mapred.child.ulimit override_mapred_child_ulimit false
Map Task Maximum Heap Size (Client Override) The maximum heap size, in bytes, of the child map processes. This number will be formatted and concatenated with 'Map Task Java Opts Base' to pass to Hadoop. Will override value in client configuration. override_mapred_map_task_max_heap false
Map Task Maximum Virtual Memory (KiB) (Client Override) The maximum virtual memory, in KiB, available to map tasks. Note: this must be greater than or equal to the -Xmx passed to the JavaVM via 'Map Task Java Opts', or else the VM might not start. This takes precedence over the generic 'mapred.child.ulimit'. Will override value in client configuration. mapred.map.child.ulimit override_mapred_map_task_ulimit false
Reduce Task Maximum Heap Size (Client Override) The maximum heap size, in bytes, of the child reduce processes. This number will be formatted and concatenated with 'Reduce Task Java Opts Base' to pass to Hadoop. Will override value in client configuration. override_mapred_reduce_task_max_heap false
Reduce Task Maximum Virtual Memory (KiB) (Client Override) The maximum virtual memory, in KiB, available to reduce tasks. Note: this must be greater than or equal to the -Xmx passed to the JavaVM via 'Map Task Java Opts', or else the VM might not start. This takes precedence over the generic 'mapred.child.ulimit'. Will override value in client configuration. mapred.reduce.child.ulimit override_mapred_reduce_task_ulimit false
Cgroup CPU Shares Number of CPU shares to assign to this role. The greater the number of shares, the larger the share of the host's CPUs that will be given to this role when the host experiences CPU contention. Must be between 2 and 262144. Defaults to 1024 for processes not managed by Cloudera Manager. cpu.shares 1024 rm_cpu_shares true
Cgroup I/O Weight Weight for the read I/O requests issued by this role. The greater the weight, the higher the priority of the requests when the host experiences I/O contention. Must be between 100 and 1000. Defaults to 1000 for processes not managed by Cloudera Manager. blkio.weight 500 rm_io_weight true
Cgroup Memory Hard Limit Hard memory limit to assign to this role, enforced by the Linux kernel. When the limit is reached, the kernel will reclaim pages charged to the process. If reclaiming fails, the kernel may kill the process. Both anonymous as well as page cache pages contribute to the limit. Use a value of -1 B to specify no limit. By default processes not managed by Cloudera Manager will have no limit. memory.limit_in_bytes -1 MiB rm_memory_hard_limit true
Cgroup Memory Soft Limit Soft memory limit to assign to this role, enforced by the Linux kernel. When the limit is reached, the kernel will reclaim pages charged to the process if and only if the host is facing memory pressure. If reclaiming fails, the kernel may kill the process. Both anonymous as well as page cache pages contribute to the limit. Use a value of -1 B to specify no limit. By default processes not managed by Cloudera Manager will have no limit. memory.soft_limit_in_bytes -1 MiB rm_memory_soft_limit true
Java Heap Size of TaskTracker in Bytes Maximum size for the Java Process heap memory. Passed to Java -Xmx. Measured in bytes. 1 GiB task_tracker_java_heapsize false

Security

Display Name Description Related Name Default Value Unit API Name Required
Users Banned from Job Submission Comma-separated list of users banned from submitting MapReduce jobs to this TaskTracker. Only applies when the TaskTracker is running in secure mode banned.users mapred, hdfs, bin taskcontroller_banned_users false
Task Controller Group The system group that owns the task-controller binary. This does not need to be changed unless the ownership of the binary is explicitly changed. mapreduce.tasktracker.group mapred taskcontroller_group false
Minimum User ID for Job Submission The lowest user ID (UID) that a user may have in order to submit a job to this TaskTracker. Only applies when the TaskTracker is running in secure mode min.user.id 1000 taskcontroller_min_user_id false
Page generated September 3, 2015.