Skip to content

Configuration

Ankit Nanglia edited this page May 23, 2018 · 15 revisions

This section explains Kronos configuration.

Task Definition

A task definition is composed of a set of attributes some required by the Kronos and some required by the task itself during execution.

KEY DESCRIPTION TYPE DEFAULT MANDATORY
name name of the task string None yes
group name of the task group that includes all its dependent tasks string default no
type type of task string None yes
isEnabled enable/ disable a task, disabled tasks won’t be picked by the Kronos for execution boolean true no
schedule cron string representing the task schedule string None yes
maxExecutionTime max allowed time for the task to finish execution string 1d no
timeoutPolicy policy to apply in case of timeout string None no
dependsOn list of tasks current task depends on list None no
properties properties required by the task handler to execute the task map None no

name: Identifies the task within a group.

group: name of the group that the task belongs to. This is similar to a namespace and can be used to isolate tasks across workflows. This feature is useful in a multi-tenant deployment where the workflow definition is same across tenants. Defining the group name same as the tenant id or name allows the same workflow definition to be re-used across tenants avoiding the need for separate deployment of Kronos across tenants.

type: Tasks of the same type are executed by the same handler. The task type is associated with a handler as part of the handler configuration.

isEnabled: By default enabled, can be turned off to remove the definition out of the Kronos. Note that if a task is disabled, all other tasks that depend on this task are marked failed.

schedule: Defines the time at which the task is to be scheduled for execution, represented as a cron string. One can use cronmaker to build simple cron string. For more details on supported cron string format check out this link.

maxExecutionTime The user can define a maximum time a task is allowed to take for its execution. Post the defined time, the task is marked as failed and if a timeout policy is configured then it is executed. Note, the task execution is not affected either by this configuration or the timeout policy.

timeoutPolicy The user can specify a policy to execute on task timeout. A task is said to be timed out if it fails to respond back within the specified max execution time.

dependsOn Defines a list of tasks it depends on. The dependant tasks are identified by their names. Dependency between two task is defined as -

KEY DESCRIPTION TYPE DEFAULT MANDATORY
name name of the task it depends on string None yes
duration look for the tasks with the specified name in the specified duration string None yes
mode specify if it depends on the first, last or all tasks found in the duration period enum all no

mode can be

  • first which means the first instance in the specified duration
  • last means the last instance in the specified duration
  • all means all the instance in the specified duration.

Duration can be defined as 5s/m/h/d, where m stand for minutes, h for hour, s for seconds and d for days. If none is passed it is assumed to be in milliseconds.

Sample task definitions where weekly aggregation task depends on daily and hourly aggregation task.

- name: Hourly Aggregation
  # task type
  type: aggregation
  schedule: 0 0 0/1 1/1 * ? *
  properties:
    key: value
    ...

- name: Daily Aggregation
  # task type
  type: aggregation
  schedule: 0 0 0 1/1 * ? *
  properties:
    key: value
    ...

- name: Weekly Aggregation
  # task type
  type: aggregation
  schedule: 0 0 0 ? * SUN *
  dependsOn:
    - name: Hourly Aggregation
      duration: 7d
      mode: all
  dependsOn:
    - name: Daily Aggregation
      duration: 7d
      mode: all
    ...
  properties:
    key: value
    ...

Note: These YAML based task definitions are read by the file based task reader extension.

properties

Additional properties might be required by the handler to execute this task. These properties will be passed on to the handler as part of task properties. Properties is a map of key-value pairs, where the key is string and value can be any object.

Task Definition Reader

The user can define custom reader to read task definitions from an external source. The user can define multiple task definition readers. The task definition readers are configured in the main configuration file app.yaml under readerConfig section. The user can also define a custom reader.

KEY DESCRIPTION TYPE DEFAULT MANDATORY
readerClass fully qualified class name of the TaskDefinitionReader implementation string None yes
schedule A cron schedule to refresh the task definitions from the source string None yes
config additional configuration required by the reader during instantiation objectnode None no

Sample configuration

readerConfig:
  # name to be assigned to the reader
  filereader:
    # fully qualified classname of the TaskDefinitionReader.java implementation
    readerClass: com.cognitree.kronos.scheduler.readers.FileTaskDefinitionReader
    # additional configuration
    config:
      source: task-definitions.yaml
    # interval in cron to look for changes in the task definition file
    schedule: 0 0/1 * 1/1 * ? *

Task Handler

For each type of task, a handler needs to be configured to execute the task. A handler is configured in the main configuration file app.yaml under handlerConfig section. The user can also define a custom handler.

KEY DESCRIPTION TYPE DEFAULT MANDATORY
handlerClass fully qualified class name of the TaskHandler implementation string None yes
config additional configuration required by the handler during instantiation objectnode None no
maxExecutionTime max allowed time for the task to finish execution if not explicitly specified by task string 1d no
timeoutPolicy policy to apply in case of timeout if not explicitly specified by task string None no
maxParallelTasks max number of tasks allowed to be executed in parallel string max cores/2 no
# configure task handlers
handlerConfig:
  # type of task to be handled by handler
  shellCommand:
    # handler implementation used to handle task of type shellCommand
    handlerClass: com.cognitree.kronos.executor.handlers.ShellCommandHandler
    # max execution time per task
    maxExecutionTime: 10m
    # max parallel tasks handler is allowed to execute at any point of time
  default:
    handlerClass: com.cognitree.kronos.executor.handlers.DefaultHandler
    # max parallel tasks handler is allowed to execute at any point of time
    maxParallelTasks: 8

In the above sample, Shell Command Handler is configured to handle tasks of type shellCommand. The user can also configure a default handler for the task type(s) that do not have a handler defined here defined as default.

Note: default is a special keyword used to define default handler.

Message Bus

Message Bus is used by scheduler and executor to exchange messages. A message bus (Producer/ Consumer) is configured for task and task status in app.yaml. The user can provide a custom implementation of message bus.

Producer

KEY DESCRIPTION TYPE DEFAULT MANDATORY
producerClass fully qualified class name of the Producer implementation string None yes
config additional configuration required by the producer during instantiation objectnode None no

Consumer

KEY DESCRIPTION TYPE DEFAULT MANDATORY
consumerClass fully qualified class name of the Consumer implementation string None yes
config additional configuration required by the consumer during instantiation objectnode None no
# configure message bus
## configure task producer
taskProducerConfig:
  producerClass: com.cognitree.kronos.queue.producer.InMemoryTaskProducer

## configure task consumer
taskConsumerConfig:
  consumerClass: com.cognitree.kronos.queue.consumer.InMemoryTaskConsumer
  config:
    pollInterval: 10s # time interval to poll tasks from task queue for execution

## configure task status producer
taskStatusProducerConfig:
  producerClass: com.cognitree.kronos.queue.producer.InMemoryStatusProducer

## configure task status consumer
taskStatusConsumerConfig:
  consumerClass: com.cognitree.kronos.queue.consumer.InMemoryStatusConsumer
  config:
    pollInterval: 10s # time interval to poll tasks status from status queue

In the above sample, an in-memory message bus is configured.

Task Store

A task store is used to store the current state of the Kronos application. A task store is configured in the main configuration file app.yaml file under storeProvider section. The user can provide a custom implementation of task store.

KEY DESCRIPTION TYPE DEFAULT MANDATORY
taskStoreClass fully qualified class name of the TaskStore implementation string None yes
config additional configuration required by the task store during instantiation objectnode None no
taskStoreConfig:
  taskStoreClass: com.cognitree.kronos.store.SQLiteTaskStore
  config:
    connectionURL: jdbc:sqlite:tasks.db
    username: SA
    password: password

In the above sample, a sqlite task store is configured.

Timeout Policy

Define policies to apply in case of task timeout. A timeout policy is configured in the main configuration file app.yaml file under timeoutPolicyConfig section. The user can provide implementation of timeout policy.

KEY DESCRIPTION TYPE DEFAULT MANDATORY
policyClass fully qualified class name of the TimeoutPolicy implementation string None yes
config additional configuration required by the timeout policy during instantiation objectnode None no
timeoutPolicyConfig:
  notifyOnTimeout:
    policyClass: com.cognitree.examples.tasks.scheduler.policies.NotifyOnTimeoutPolicy
    config:
      smtp.host: "smtp.gmail.com"
      smtp.port: 587
      sender.name:
      sender.email:
      sender.password:
      email.subject: "Task Timeout notification"
      recipient.name:
      recipient.email:

In the above sample, notifyOnTimeout is the policy identifier that is to be used while defining the task definition and/or task handler as the timeout policy.

Task Purge Interval

Periodically, tasks older than the specified interval and status in one of the final state are purged from memory. The task purge interval is configured in the main configuration file app.yaml under taskPurgeInterval section.

KEY DESCRIPTION TYPE DEFAULT MANDATORY
taskPurgeInterval purge interval (1d, 1h 15m etc) string 1d no
taskPurgeInterval: 6h

In the above sample, the task purge interval period is configured as 6 hours.