Awesome

DEPRECATED

This repository is now longer maintained. Please use mesos_exporter.

mesos-task-exporter

Export metrics of tasks running on Apache Mesos for Prometheus consumption.

Features

• Auto-discover of new slaves
• Auto-discover of new tasks
• Stateless - can run anywhere
• Framework-independent

Metrics

Task

Exported metrics

• mesos_task_cpus_limit
• mesos_task_cpus_system_time_seconds
• mesos_task_cpus_user_time_seconds
• mesos_task_mem_limit_bytes
• mesos_task_mem_rss_bytes

Labels

Every metric has the following labels attached to it:

• executor_id - The unique ID of the executor
• framework - The name of the framework that spawned the task
• slave_pid - The PID of the Mesos slave that the task is running on as exposed by the /master/state.json endpoint of the Mesos master
• task - The name of the task as in the Mesos UI

Examples

Chronos:

mesos_task_cpus_system_time_seconds{executor_id="ct:1426247880000:0:examplejob:",framework="chronos-2.3.2_mesos-0.20.1-SNAPSHOT",slave_pid="slave(1)@10.168.1.11:5051",task="ChronosTask:examplejob"} 0.02

Marathon:

mesos_task_cpus_system_time_seconds{executor_id="com_example_redis.b8f17462-c96c-11e4-b9ff-56847afe9799",framework="marathon",slave_pid="slave(1)@10.168.1.11:5051",task="redis.example.com"} 10.71

Global task stats

Exported metrics

mesos_tasks

Labels

status

Example

mesos_tasks{status="failed"} 0
mesos_tasks{status="finished"} 1
mesos_tasks{status="killed"} 0
mesos_tasks{status="lost"} 0
mesos_tasks{status="staged"} 2
mesos_tasks{status="started"} 0

Resources advertised by a Mesos slave

Exported metrics

mesos_slave_resources

Labels

• pid - The unqiue PID of the slave in the Mesos cluster
• resource - The name of the resource

Example

mesos_slave_resources{pid="slave(1)@10.168.1.10:5051",resource="cpus"} 2
mesos_slave_resources{pid="slave(1)@10.168.1.10:5051",resource="disk"} 35164
mesos_slave_resources{pid="slave(1)@10.168.1.10:5051",resource="mem"} 748

Resources used by a framework

Exported metrics

mesos_framework_resources

Labels

• name - The name of the framework
• resource - The name of the resource
• type - The type of the metric. Currently only used is implemented.

Example

mesos_framework_resources{name="marathon",resource="cpus",type="used"} 0.2
mesos_framework_resources{name="marathon",resource="disk",type="used"} 0
mesos_framework_resources{name="marathon",resource="mem",type="used"} 256

Configuration

$ ./mesos-task-exporter -h
Usage of ./mesos-task-exporter:
  -exporter.address=":55555": Address of the exporter
  -exporter.endpoint="/metrics": Path where metrics are served
  -log.level="info": Log level
  -mesos.master-pollinterval=15s: Interval to poll the Mesos master leader for new slaves
  -mesos.masters="http://localhost:5050": A list of Mesos masters separated by commas
  -mesos.slave-pollinterval=15s: Interval to poll a Mesos slave for stats of tasks

# prometheus.yml
scrape_configs:
  - job_name: mesos_task_exporter
    scrape_interval: 15s

    target_groups:
      - targets: ['localhost:55555/metrics']

Deployment

Run mesos-task-exporter as a Docker container:

docker run -p 55555:55555 -e "MESOS_MASTERS=..." wandhydrant/mesos-task-exporter