config

Manages configuration values for the MapR cluster.

Configuration Fields

The following fields are configurable.

Field

Default Value

Description

cldb.balancer.disk.max.switches.in.nodes.percentage 10 The maximum number of containers that can be balanced in parallel by the disk balancer. The value is a percentage of the number of nodes in the system.
cldb.balancer.disk.paused 1 Enables or disables the Disk Balancer.
cldb.balancer.disk.sleep.interval.sec 2 * 60 The sleep interval (in seconds) between two successive runs of the Disk Balancer.
cldb.balancer.disk.threshold.percentage 70 Percentage of used space that causes containers in a storage pool to be distributed across other less used storage pools.
cldb.balancer.logging 0 Disables (0) or enables (1) the logging of messages in Disk Balancer and Role Balancer.
cldb.balancer.role.max.switches.in.nodes.percentage 10 The percentage (of the number of nodes in the system) to use to determine the maximum number of containers whose roles (Masters and Tails) will be balanced in parallel by the Role Balancer.

For example, suppose there are 500 nodes and this value is 10(%). The number of containers whose roles are balanced in parallel is (10/100)*500=50.

cldb.balancer.role.paused 1 Enables (1) or Disables (0) the Role Balancer.
cldb.balancer.role.sleep.interval.sec 15 * 60 The sleep interval (in seconds) between two successive runs of the Role Balancer.
cldb.balancer.startup.interval.sec 30 * 60 The initial startup delay of the Role Balancer for existing clusters.
cldb.cluster.almost.full.percentage 90 The percentage at which the CLUSTER_ALARM_CLUSTER_ALMOST_FULL alarm is triggered.
cldb.container.alloc.selector.algo 0 The allocation algorithm to use when creating new containers. Values can be:
  • 1 - indicates RoundRobin algorithm. If selected, containers are allocated across nodes in a topology in a roundrobin fashion.
  • 2 - indicates Randomized algorithm. If selected, containers are allocated across nodes in a randomized way.
  • 0 - indicates RoundRobin algorithm if the number of nodes is less than or equal to 100, Randomized algorithm otherwise.
cldb.container.assign.buffer.sizemb 1 * 1024 The size of the container that should be used as a buffer. When allocating a new container, this size is deducted from the maximum container size.
Note: When cldb.container.sizemb value is modified, check and update the value of cldb.container.assign.buffer.sizemb to prevent new containers from being created when existing containers are not full.
cldb.container.create.diskfull.threshold 85

The percentage of space on a fileserver to use to classify the fileserver as FULL.

cldb.container.sizemb 32 * 1024

The maximum size for containers. This is a soft limit.

Note: When cldb.container.sizemb value is modified, check and update the value of cldb.container.assign.buffer.sizemb to prevent new containers from being created when existing containers are not full.
cldb.default.chunk.sizemb 256

The size of each chunk that make up a file in the MapR file system.

cldb.default.volume.topology /data

The default topology for new volumes.

cldb.dialhome.metrics.file.rotation.period 365

The retention period of the files (in days) used to record Dialhome metrics. Files that are past their retention period are automatically deleted.

cldb.fs.mark.rereplicate.sec 60 * 60

The number of seconds a node can fail to heartbeat before it is considered dead. Once a node is considered dead, the CLDB re-replicates any data contained on the node.

cldb.fs.reregistration.wait.time 15 The amount of time (in minutes) to wait before checking for inactive nodes.
Note: Reduce the value to raise alarm without delay after CLDB failover. To avoid spurious alarms, do not reduce value below 5 (minutes).
cldb.log.fileserver.timeskew.interval.mins 60

The frequency in minutes at which CLDB should log messages about fileserver's time skew.

cldb.max.parallel.resyncs.star 3

The number of container replicas that can resync in parallel from the source for low_latency (star-replicated) volumes.

cldb.mfs.heartbeat.timeout.multiple 10

Specifies a heartbeat timeout multiple. For small clusters, the heartbeat interval is 1 second and the multiple is 10 by default, which makes the heartbeat timeout 10 seconds.

cldb.min.fileservers 1

The number of fileservers hosting the CLDB volume required for the master CLDB to complete the bootstrap process.

cldb.replication.manager.critical.paused 0

Disables (0) or enables (1) the processing of critically under-replicated containers. If enabled, the critically under-replicated containers are processed on a priority basis to increase the number of copies.

cldb.replication.manager.max.resyncs.in.nodes.percentage 1200

The number of containers that can be replicated in parallel, expressed as a percentage of the number of active nodes. If the value is 1200, the number of containers that can be replicated is 12 times the number of active nodes.

cldb.replication.manager.over.paused 0

Disables (0) or enables (1) the processing of over-replicated containers. Over-replicated containers are processed to delete extra copies, which is when the number of copies is more than the desired replication factor.

cldb.replication.manager.start.mins 15

The delay between CLDB startup and replication manager startup, to allow all nodes to register and heartbeat

cldb.replication.max.in.transit.containers.per.sp 4

The maximum number of containers that can be in transit on an SP. Containers that are serving either as the source or destination of a resync operation are considered as being in ‘transit’.

cldb.replication.sleep.interval.sec 15

The sleep duration (in seconds) between consecutive runs of the Replication Manager.

cldb.replication.tablescan.interval.sec 2 * 60

The sleep duration (in seconds) between consecutive runs of the Replication Scanner. While the Replication Scanner bucketizes containers into different classes, the Manager thread either replicates or removes additional copies.

cldb.security.user.ticket.duration.seconds 1209600 The length of time before the user ticket (generated using maprlogin password command) expires.
cldb.security.user.ticket.max.duration.seconds 2592000 The maximum amount of time allowed for the user ticket (generated using the maprlogin password command).
cldb.security.user.ticket.renew.duration.seconds 2592000 The length of time to renew the user ticket (generated using maprlogin password command).
cldb.security.user.ticket.renew.max.duration.seconds 7776000 The maximum duration (7776000, by default) allowed for a user ticket (generated using maprlogin password command) renewal.
cldb.topology.almost.full.percentage 90 The threshold percentage that is used to raise alarms when the used space on the nodes of a topology exceed a certain percentage of total space.
cldb.volume.epoch The starting epoch of a new Container. Epoch are used internally in the election of the master container etc.

cldb.volumes.namespace.default.min.replication

2

The minimum replication factor for the name container. Containers with fewer copies than this value are replicated on a priority basis.

cldb.volumes.namespace.default.replication

3

The desired replication factor for the name container.

mapr.fs.nocompression "bz2,gz,tgz,tbz2, zip,z,Z,mp3,jpg, jpeg,mpg,mpeg,avi, gif,png,lzo,jar" The file types that should not be compressed. See File Extensions of Compressed Files.
mapr.fs.permissions.supergroup root The super group of the MapR file system layer.
mapr.fs.permissions.superuser mapr The super user of the MapR file system layer.
mapr.targetversion The configuration variable to set the current version of mapr distribution. Failing to set this variable on an upgrade will cause alarms to be missed when all the nodes in a cluster are not at the same version of the software.
mfs.db.parallel.copyregions 4 The number of parallel copy regions per MFS instance. Setting this field to a larger value increases the parallelism for data transfers during index updates, CDC propagation, and table replication. This increases the transfer rate and reduces the initial synchronization time, but uses more system resources. The latter may impact the response time and performance of applications reading data from the same nodes.
mfs.high.memory.alarm.threshold 110 (percentage of allocated memory) Threshold for raising NODE_ALARM_HIGH_MFS_MEMORY alarm.
mfs.feature.db.json.support 1 for new MapR installations, 0 for upgraded MapR installations. Disables (0) or enables (1) MapR streams and support in MapR-DB for JSON documents and tables.
mfs.feature.devicefile.support 1 Defines whether named pipes can be used over NFS.
pernode.numcntrs.alarm.thr 50000 The maximum number of RW containers on each node beyond which performance may not be optimal. The optimal number for RW and snapshot containers combined is 10 times the value of this parameter.