Product Knowledge Base

Space shortcuts

Page tree

Installing the JobScheduler Universal Agent

Prerequisites

  • A Java Runtime Environment starting from version 1.8 is required.

Preparation

  • Choose the JobScheduler Universal Agent archive for the target system from the JobScheduler Downloads page.
    • For Unix systems:
    • For Windows systems:
      • A .zip archive is available that can be used for:
        • manual installation on a small number of computers.
          (Described in the current article.)
        • the installation of a large number of Agents using third party deployment tools.
      • An installer that can be used for manual installation a small number of computers.
        (Described in detail in the JobScheduler Universal Agent - Installation with Windows Installer article.)

Installation

    • Unzip the downloaded file to an arbitrary directory.
    • Directory structure (only files and folders directly relevant ):
      • bin
        • jobscheduler_agent.cmd
          • The start script for Windows platforms.
        • jobscheduler_agent.sh
          • The start script for Unix platforms.
        • jobscheduler_agent_instance.sh-example
      • lib
        • The directory for Java libraries.
        • Configure the settings in log4j.xml file if you want to adjust the log output format.
      • var_4445
      • service (for Windows)
    • On Windows Systems:
      • You have to change the directory permissions for the .\logs and .\service directories if you have extracted the JobScheduler Agent under e.g. C:\Program Files.
        • This step is not required if you extracted the JobScheduler Agent to e.g. C:\ProgamData.

        • Start a command prompt with elevated administrator rights and type e.g.

          set full acces for "Users" on .\service and .\logs directory
          cd "path\to\JobScheduler Agent installation"
icacls "service" /L /grant *S-1-5-32-545:(OI)(CI)F
icacls "logs" /L /grant *S-1-5-32-545:(OI)(CI)F 
    • If multiple instances are configured then every instance must have its own ./var_4445 data directory (e.g. ./var_<port of the instance>)

Update of a JobScheduler Universal Agent

  • Preparations for an update
    • Stop all JobScheduler Universal Agent instances
    • Remove the ./lib directory
  • Unzip the downloaded file to the installation directory
  • If you use the Windows installer of JobScheduler Universal Agent then the ./lib directory is updated automatically during the setup. It is not necessary to remove the ./lib directory before you start the installer.

Running the JobScheduler Universal Agent

  • SOS does not recommend running the JobScheduler Agent as root (for Unix) or as Administrator (for Windows).
  • Instead the user account should be used that jobs are executed for. Should jobs be executed for a number of user accounts then consider the chapter Running multiple instances of JobScheduler Universal Agent.

Usage

Running the start script without parameters shows the usage clause:

Usage for Unix
Usage: jobscheduler_agent.sh command [options] [web services]
  command:
    start           [options]
    start_docker    [options]
    stop            [options]
    abort           [options]
    restart         [options]
    status          [options] [web services]
    start           [options]
 options:
    -http-port=<[hostname or ip address:]number>   | to listen to a specific host name or ip address if it is specified; default port: 4445, 
    -https-port=<[hostname or ip address:]number>  | 
    -data-directory=<location of data directory>   | default: ./var_4445
    -timeout=<number>                              | in seconds; only for stop and restart
    -kill-script=<location of kill script>         | default: "./bin/jobscheduler_agent_kill_task.sh"; only for start
    -java-options=<java options>                   | default: -Xms100m; see https://kb.sos-berlin.com/x/aIC9
    -job-java-options=<java options>               | see https://kb.sos-berlin.com/x/aIC9
 web services:
    overwiew                                       | default
    task                                           | running JobScheduler Agent tasks
    tunnel                                         | tunnels summary
    tunnel/                                        | list of tunnels
    command                                        | running command summary
    command/                                       | list of running commands (watching file_order_source)
Usage for Windows
Usage: jobscheduler_agent.cmd command [options] [web services]
  command:
    start           [options]
    stop            [options]
    abort           [options]
    restart         [options]
    status          [options] [web services]
    debug           [options]
    kill            [options]
    install-service [options]
    remove-service  [options]
    start-service   [options]
 options:
    -http-port=<[hostname or ip address:]number>   | to listen to a specific host name or ip address if it is specified; default port: 4445, 
    -https-port=<[hostname or ip address:]number>  | 
    -data-directory=<location of data directory>   | default: ./var_4445
    -timeout=<number>                      | in seconds; only for stop and restart
    -kill-script=<location of kill script> | default: ".\bin\jobscheduler_agent_kill_task.cmd"; only for start, debug
    -java-options=<java options>           | default: -Xms100m; see https://kb.sos-berlin.com/x/aIC9
    -job-java-options=<java options>       | see https://kb.sos-berlin.com/x/aIC9
 web services:
    overwiew                               | default
    task                                   | running JobScheduler Agent tasks
    tunnel                                 | tunnels summary
    tunnel/                                | list of tunnels
    command                                | running command summary
    command/                               | list of running commands (watching file_order_source)

Command Line Options

  • -http-port=<[hostname or ip address:]number>
    •  is the HTTP port that the Agent is listening to in order to receive requests from a JobScheduler Master:
      • jobscheduler_agent.cmd|sh command -http-port=####
      • where #### is the numeric port.
      • This option can be also used for indicating which network interfaces the JobScheduler Agent should listen to if you specify a hostname or ip address in addition
    • Without this option being used the port defaults to 4445 and the JobScheduler Agent listens to all available network interfaces.
    • Should you want to specify a port then the following precedence applies:
      • First precedence: command line option
      • Second precedence: environment variable SCHEDULER_HTTP_PORT (see below)
      • Third precedence: use of default value
  • -https-port=<[hostname or ip address:]number>
    •  is the HTTPS port that the Agent is listening to in order to receive requests from a JobScheduler Master:
      • jobscheduler_agent.cmd|sh command -https-port=####
      • where #### is the numeric port.
      • This option can be also used for indicating which network interfaces the JobScheduler Agent should listen to if you specify a hostname or ip address in addition
    • Without this option being used the port defaults to 4445and the JobScheduler Agent listens to all available network interfaces.
    • Should you want to specify a port then the following precedence applies:
      • First precedence: command line option
      • Second precedence: environment variable SCHEDULER_HTTP_PORT (see below)
      • Third precedence: use of default value
  • -data-directory=<number>
    • Location of the data directory.
    • It has to be unique over all JobScheduler Universal Agent instances
    • Should you want to specify a data directory then the following precedence applies:
      • First precedence: command line option
      • Second precedence: environment variable SCHEDULER_DATA (see below)
      • Third precedence: use of default value (=SCHEDULER_HOME/var_SCHEDULER_HTTP_PORT)
  • -timeout=<number>
    • This option can be used to specify the number of seconds that the Agent will wait for tasks to stop. 
    • This option can be applied for stop and restart commands.
    • The Agent sends a SIGTERM signal to the taksk and having reached the timeout a SIGKILL signal will be sent to stop any tasks immediately.
  • -kill-script=<location of kill script>

    • The kill scripts provide the functionality to kill a task and it's child processes.

    • Two kill scripts are provided

      • ./bin/jobscheduler_agent_kill_task.sh for Unix as default

      • .\bin\jobscheduler_agent_kill_task.cmd for Windows as default

    • This option can be used to specify the location of a different "kill script" if necessary

    • Should you want to specify a different "kill script" then the following precedence applies:

      • First precedence: command line option
      • Second precedence: environment variable SCHEDULER_KILL_SCRIPT (see below)
      • Third precedence: use of default value
  • -java-options=<java options>

    • With Java 1.8 the initial memory allocation has changed, for details see How to manage the Java heap space.

    • This option can be used to apply Java options for the JobScheduler Universal Agent, e.g. the memory settings.

    • Without this option being used the Java options default to '-Xms100m'.

    • Should you want to specify the Java options then the following precedence applies:

      • First precedence: command line option
      • Second precedence: environment variable JAVA_OPTIONS (see below)
      • Third precedence: use of default value
  • -job-java-options=<java options>

    • Without this option being used the Java options for each job which is started by the JobScheduler Universal Agent.

    • Should you want to specify the Java options for the jobs then the following precedence applies:

      • First precedence: command line option
      • Second precedence: environment variable SCHEDULER_JOB_JAVA_OPTIONS (see below)

Start the Agent

jobscheduler_agent.cmd|sh start [options]

Stop the Agent

jobscheduler_agent.cmd|sh stop [options]

This command will safely terminate the Agent (recommended).

  • The Agent waits for running processes to be completed.
  • Any child processes will be killed by the Agent.


jobscheduler_agent.cmd|sh abort [options]

The Agent process is terminated immediately.

  • Any running tasks and child processes are killed immediately with a SIGKILL signal.
  • Should task have used resources such as database connections then they will not be properly closed.


jobscheduler_agent.cmd|sh kill [options]

The Agent process is killed.

  • This corresponds to sending SIGKILL with a kill command.
  • Should jobs be running that started detached child processes then it is not guaranteed that child processes will be killed.

Restart the Agent

jobscheduler_agent.cmd|sh restart [options]

Query the Agent Status

jobscheduler_agent.cmd|sh status [options]

Should the Agent be up and running then this command will result in some output such as:

isTerminating: false
system:
  hostname: agenthost
  mxBeans:
    operatingSystem:
      processCpuLoad: 2.5630713121704744E-5
      availableProcessors: 4
      freePhysicalMemorySize: 311668736
      systemCpuLoad: 0.046373903924522855
      committedVirtualMemorySize: 4475301888
      totalPhysicalMemorySize: 3155517440
java:
  systemProperties:
    java.vendor: Oracle Corporation
    os.arch: amd64
    os.version: 2.6.32-220.17.1.el6.x86_64
    os.name: Linux
    java.version: 1.8.0_31
version: 1.10.0-SNAPSHOT (6956c56a535d15fcf659f293c42d22dcf92e9e12 2015-07-15 21:23:24+02:00)
startedAt: '2015-07-17T08:38:30.516Z'
totalTaskCount: 21
currentTaskCount: 0


Should the Agent not be running then some output is provided such as:

ERROR: spray.can.Http$ConnectionAttemptFailedException: Connection attempt to localhost:4445 failed
...JobScheduler Agent(4445) not started!

Monitoring Agents

See the How to perform active checks with a System Monitor such as Nagios/op5 article for information about configuring an automated status check of Agents for monitoring purposes.

Master/Agent Compatibility

The Universal Agent will raise an error if it is contacted by a Master JobScheduler older than 1.10.

JS-1509 - Getting issue details... STATUS

Working Directory for Jobs

The SCHEDULER_WORK_DIR environment variable can be used to change the jobs working directory from the default value of $SCHEDULER_HOME to a value such as $HOME (Unix) or %USEERPROFILE% (Windows).

JS-1527 - Getting issue details... STATUS

Change Management References - Agent Usage

T Key Linked Issues Fix Version/s Status P Summary Updated
Refresh

Windows Service Interface: Usage

The following information applies to batch installation on Windows systems. For installation with a GUI and user dialog see JobScheduler Universal Agent - Installation with Windows Installer.

The JobScheduler Universal Agent is operable as a Windows Service. The start script of the JobScheduler Universal Agent allows to install/remove the Windows Service.

JS-1325 - Getting issue details... STATUS

Install the Windows Service

jobscheduler_agent.cmd install-service [-http-port=<number>] [-ip-address=<hostname or ip address>]

This command installs the Windows Service. After the installation you find the Windows Service with the name SOS JobScheduler Agent -port=<number> in the Services Manager Console. The Windows service uses the "local system" account.

During the service installation it tries to copy the executable file for the Windows Service to the .\service directory. This operation could fail with the error "Access denied" if you have extracted the JobScheduler Universal Agent to e.g. C:\Program Files\. In this case you can change the permissions of the .\service directory or open the command prompt with elevated administrator rights and execute the above command once more.

Start the Windows Service

jobscheduler_agent.cmd start-service [-http-port=<number>]

This command starts the Windows Service with the name SOS JobScheduler Agent -port=<number>.

The stop command contains more than a simple stop-service command: the stop command checks whether the Agent was started through the CLI or as a Windows Service and stops the Agent accordingly. Therefore there is no stop-service command.

Remove the Windows Service

jobscheduler_agent.cmd remove-service [-http-port=<number>]

This command removes the Windows Service. After executing this command you will not find the Windows Service with the name SOS JobScheduler Agent -port=<number> in the Services Manager Console any longer.

During removal of the service it tries to remove the executable file of the Windows Service from the .\service directory. This operation could fail with the error "Access denied" if you have extracted the JobScheduler Agent to e.g. C:\Program Files\. In this case you can change the permissions of the .\service directory or open the command prompt with elevated administrator rights and execute the above command once more.

Configure the Windows Service

After the installation of the Windows Service you will find the .\service\sos_jobscheduler_agent_<http-port>w.exe file. Start this program to configure the Windows Service.

For example goto the "Startup" tab
to modifiy the start parameter

Change Management References - Windows Service

T Key Linked Issues Fix Version/s Status P Summary Updated
Refresh

Logging

JS-1393 - Getting issue details... STATUS

JS-1497 - Getting issue details... STATUS

  • Log File
    • On startup the Agent creates a log file in the directory that is pointed to by the environment variable SCHEDULER_LOG_DIR or in the var_4445/logs subdirectory of the Agent installation directory.
    • Log file names are created from a prefix and the port used by the Agent like this:
      • jobscheduler_agent_4445.log
    • Log files are rotated for each day (see ./lib/log4j.xml) for which job activities occur.
    • Rotated log files are assigned file names like this: jobscheduler_agent_4445.log.<yyyy-MM-dd>
    • For days where the JobScheduler Agent has no jobs to execute no log rotation will be performed.
  • PID File
    • On startup the Agent creates a PID file in the directory that is pointed to by the environment variable SCHEDULER_PID_FILE_DIR or in the log directory. The PID file contains the Process ID of the system process that the Agent is running in.
    • The PID file is used in order to prevent the Agent to be started twice with the same settings and it can be used for shutdown scripts that require the PID to terminate the process.
    • PID file names are created from a prefix and from the port used by the Agent like this:
      • jobscheduler_agent_4445.pid
  • Further References

Change Management References - Logging

T Key Linked Issues Fix Version/s Status P Summary Updated
Refresh

Environment Variables

JS-1461 - Getting issue details... STATUS

JS-1443 - Getting issue details... STATUS

JS-1442 - Getting issue details... STATUS

The following environment variables can be used:

  • JAVA_HOME
    • points to the location of the Java Runtime Environment (JRE).
    • Without setting this environment variable Java will be used from the location specified by the system path.
    • Please consider that JAVA_HOME does not point to the location of a JDK but to a JRE directory where the bin/java executable resides.
  • JAVA_OPTIONS
    • sets Java options, e.g the Java memory settings for the JobScheduler Universal Agent.
    • Without setting this environment variable the Java options defaults to '-Xms100m'.
  • SCHEDULER_JOB_JAVA_OPTIONS
    • sets Java options for each job which is started by the JobScheduler Universal Agent.
  • SCHEDULER_HOME
    • points to the directory where the JobScheduler Agent has been installed.
    • Without setting this environment variable the default value is the parent directory of the start script.
    • Should you want to start the Agent from a directory different to the Agent installation directory, e.g. by copying the start script to some other location, then this environment variable has to be set in order to locate the JobScheduler Agent installation directory.
  • SCHEDULER_DATA
    • points to the directory where the JobScheduler Agent has its data directory.
    • Without setting this environment variable the default value is SCHEDULER_HOME/var_SCHEDULER_HTTP_PORT.
  • SCHEDULER_HTTP_PORT
    • sets the http port that the JobScheduler Agent is listening to.
    • indicates which network interfaces the JobScheduler Agent should listen to if a host or ip address specified.
    • if only a port number is specified then the JobScheduler Agent listens to all available network interfaces via http.
    • Without setting this environment variable the port defaults to 4445.
  • SCHEDULER_HTTPS_PORT
    • sets the https port that the JobScheduler Agent is listening to.
    • indicates which network interfaces the JobScheduler Agent should listen to if a host or ip address specified.
    • if only a port number is specified then the JobScheduler Agent listens to all available network interfaces via https.
    • Without setting this environment variable the https protocol doesn't use.
  • SCHEDULER_USER
    • sets the user account that the JobScheduler Agent is operated for. This includes running jobs with the permissions of the specified user.
    • This setting is available for Unix systems only. For Windows systems the user account that runs the start script is used.
    • Without setting this environment variable the user acount that runs the start script is used.
    • This setting can be used when running the Agent start script in system start-up and shutdown configurations that are executed by root, e.g. in /etc/init.d or corresponding locations.
  • SCHEDULER_LOG_DIR
    • sets the directory where the JobScheduler Agent log file is created.
    • This setting defaults to the directory logs in the Agent installation directory.
    • For Windows systems for which the Agent is installed in the program directory that is pointed to by the %ProgramFiles% environment variable it is recommended not to use the default setting. Instead specify a different path via the SCHEDULER_LOG_DIR environment variable, e.g. some location in the data directory that is pointed to by the %ProgramData% environment variable.
  • SCHEDULER_WORK_DIR
    • sets the working directory for the jobs started by the JobScheduler Agent, e.g. ${HOME} or %USERPROFILE%.
    • This setting defaults to the SCHEDULER_HOME.
  • SCHEDULER_KILL_SCRIPT
    • sets the location of a "kill script" if necessary.

    • The kill scripts provide the functionality to kill a task and it's child processes.

    • Two kill scripts are provided

      • ./bin/jobscheduler_agent_kill_task.sh for Unix as default

      • .\bin\jobscheduler_agent_kill_task.cmd for Windows as default

  • SCHEDULER_PID_FILE_DIR
    • sets the directory where the JobScheduler Agent pid file is created.
    • This setting defaults to the directory that is specified with the SCHEDULER_LOG_DIR environment variable or the log directory default value.

Change Management References - Environment Variables

T Key Linked Issues Fix Version/s Status P Summary Updated
Refresh

Usage Examples

Running the JobScheduler Agent on Windows

For Windows® operating systems the location of the Java Runtime Environment and of the log directory can be specified like this:

set JAVA_HOME=%ProgramFiles%\Java\jre8
set SCHEDULER_LOG_DIR=%ProgramData%\sos-berlin.com\agent\jobscheduler_agent_1.10\logs 
"%ProgramFiles%\sos-berlin.com\agent\jobscheduler_agent_1.10\bin\jobscheduler_agent.cmd" start

Running the JobScheduler Agent on Mac OS X

For Mac® OS X the location of the Java Runtime Environment can be specified like this:

JAVA_HOME=/Library/Internet\ Plug-Ins/JavaAppletPlugin.plugin/Contents/Home
export JAVA_HOME
/Users/ap/Documents/jobscheduler_agent/bin/jobscheduler_agent.sh start

Automated Start-up and Shutdown of JobScheduler Universal Agent

  • For Unix systems the start-up and shutdown configurations apply that are executed by root, e.g. in /etc/init.d or corresponding locations.
    • Consider use of the SCHEDULER_USER environment variable to run an Agent that is started by root for a different user account.
  • For Windows systems the start-up of the Agent by installing it as a Windows Service is recommended.


Debugging

  • The Agent log level can be increased using the Agent's Apache ProcRun Demon Service Manager demon/service.
  • On Windows systems this is installed in the Agent service Folder and will have a name such as sos_jobscheduler_agent_4444w.exe where 4444 is the port the agent can be addressed over. 
  • Start the ProcRun Manager, select the Logging tab in the Manager interface and set the level to Debug. 
  • The location of the log files has already been described above. 
  • (Do not forget to set the debug level back to Info once finished.)


Running multiple instances of JobScheduler Universal Agent

Testing the JobScheduler Universal Agent Operability

A simple way to test if the installed JobScheduler Agent works as expected is to carry out a test using Process Classes. The Process Class and the corresponding job will be defined in your JobScheduler Master, which will connect to the JobScheduler Agent. SOS recommends to do the test locally to exclude connection problems, e.g. firewall settings, since the goal of the test is to see whether the JobScheduler Agent is working properly.

Process Class and Standalone Job

First create a Process Class, for example the following (find attached example: agent1.process_class.xml):

Example Process Class
<?xml version="1.0" encoding="ISO-8859-1"?>
<process_class  max_processes="10">
    <remote_schedulers >
        <remote_scheduler  remote_scheduler="http://localhost:4445"/>
    </remote_schedulers>
</process_class>

Important! When defining the remote JobScheduler URL, it is required to add 


Second, you should define a Standalone Job and associate the Process Class as configured above to the job (find an example attached: standalone_hello.job.xml):

Example Standalone Job
<?xml version="1.0" encoding="ISO-8859-1"?>

<job  process_class="agent1">
    <script  language="shell">
        <![CDATA[
echo "hello world"
        ]]>
    </script>
    <run_time />
</job>

Finally:

  • store both XML files (agent1.process_class.xml and  standalone_hello.job.xml) to the JobScheduler live folder in ./config/live.
  • start the job in JOC (start task immediately). 

Find more details and use cases from the article: How to execute Jobs and Job Chains with Agents

Logs

The easiest way to check if the test was successful is to verify the contents of the job log in JOC and to see if the job completed successfully. Otherwise an error will be reported in JOC.

For double-checking you can verify the log from the JobScheduler Universal Agent. If the JobScheduler Universal Agent is running and the steps above were correctly completed then you should see something like this in the JobScheduler Universal Agent log:

JobScheduler Universal Agent Log
2015-04-28 16:11:24.462 +0200 [DEBUG] akka.io.TcpListener - New connection accepted
......
......
2015-04-28 16:13:39.231 +0200 [DEBUG] spray.can.server.HttpServerConnection - TcpConnection terminated, stopping

Remote File Watching

To configure remote file watching, see JobScheduler Universal Agent - Remote File Watching

Behavior in the event of the Universal Agent crashing

It is important that all the tasks running on a Universal Agent are killed if the Agent should crash or otherwise terminate abnormally while executing tasks. To this end, every task that is being executed by a Universal Agent is noted by a jobscheduler_agent_watchdog.sh agent watchdog process in a kill_tasks_after_crash.sh script. This script is located in the Agent's tmp folder and tasks are dynamically added to and deleted from the script as they are started and completed. This script is dynamically created when a first task is started by the agent and deleted when no tasks are running.

If a Universal Agent crashes the Agent Watchdog will start this script which will then cause all tasks that were running when the Agent crashed to be killed.

See also