Example Command

Commands are delivered to the Agent from the Agent Controller over long polling. Agent keeps polling the AC for jobs. Jobs are delivers as json objects as following

 {
    "id": "<job-id>",
    "gid": <grid-id>,
    "nid": <node-id>,
    "name": "<command>",
    "args": {
        ...
    },
    "data": "<data-string>",
    "role": "<optional role>",
    "fanout": true or false,
}

• id: Job ID. Chosen by the caller
• gid: Grid ID
• nid: Node ID
• args: Agent speicific RUN arguments (also called runargs in some parts of the documentation).
• data (optional): Data string, will be fed to the sub-process over stdin. So it actually can be anything including serialized json data.
• role: (mutual exclusive with gid) specifies agent role that must execute this job. If provided the first agent to satisfy this role will get the job. There is a special role * which means (any agent).
• fanout: Fanout only effective when a role is provided. Basically all agents that satisfy the given role will get the job.

If role=* and fanout=true this basically means ALL agents.

args

Arguments fine-tunes the process. The arguments are interpreted by the Agent itself to control the behavior of the sub-process, unlike data which is passed unprocessed to the sub-process itself. Also called runargs in some parts of the documentation. Basic arguments support the following args:

• max_time: How long the sub-process can run, 0 for (forever) or long running process, value > 0 means kill the process of execution time takes more than max_time. If max time is set to -1 this instructs the agent to remember the job and will rerun it on agent start (useful with recurring_period).
• max_restart: max times the process will be restarted if failed, if max_restart = 0 then no restart. If the process lived for more than 5 min before it failes the counter will be reset.
• domain: domain of jumpscript (to do categorization)
• name: name of jumpscript or cmd to execute (will give all a name)
• loglevels: list of log levels to process
• loglevels_db: list of log levels to be processed by DB logger (overrides the logger defaults)
• loglevels_ac: list of log levels to be processed by AC logger (overrides the logger defaults)
• recurring_period: 0 or 100 seconds between recurring execute this cmd 0 is default means not recurring, so only once
• stats_interval: 120 means we overrule the default for this process and only monitor this porcess every 120 seconds.
• queue: If queue is set, the command will wait on a job queue for serial execution. In other words no 2 processes with the same queue name will get executed on the same agent at the same. time

execution environment

When using command execute or any of the external extensions the agent defines a set of environment variables that can be accessed by the child process:

• AGENT_HOME: CWD of the agent
• AGENT_GID: Gid of the agent
• AGENT_NID: Nid of the agent
• AGENT_CONTROLLER_NAME: Name of the agent controller that intiated this command (As defined in the agent config)
• AGENT_CONTROLLER_URL: Base url of the agent controller
• AGENT_CONTROLLER_CLIENT_CERT: Path to agent client certificates
• AGENT_CONTROLLER_CLIENT_CERT_KEY: Path to agent client certificates key

Built in commands

• execute
• get_cpu_info
• get_disk_info
• get_mem_info
• get_nic_info
• get_os_info
• killall
• ping
• restart
• get_msgs
• del_msgs