The Charmander-Scheduler offers a REST-like API for its task scheduling functionality. For this documentation
we assume that Charmander is running at its default location at 172.31.1.11:7075.
Simple ping-pong healthcheck
curl -i 172.31.1.11:7075/admin/ping
Should return a pong.
Start a Task
To start a task in Mesos you will have to pass in a task-configuration.
Example configuration for cAdvisor:
{
"id": "cadvisor",
"dockerimage": "google/cadvisor:0.6.2",
"mem": "50",
"sla": "one-per-node",
"notmetered": true,
"volumes": [
{
"host_path":"/",
"container_path":"/rootfs",
"mode":"ro"
},
{
"host_path":"/var/run",
"container_path":"/var/run",
"mode":"rw"
},
{
"host_path":"/sys",
"container_path":"/sys",
"mode":"ro"
},
{
"host_path":"/var/lib/docker",
"container_path":"/var/lib/docker/",
"mode":"ro"
}
],
"ports": [
{
"host_port":"31500",
"container_port":"8080"
}
]
}
Example configuration for stress60mb:
{
"id": "stress60",
"dockerimage": "stress",
"mem": "100",
"reshuffleable": true,
"nodetype": "lab",
"arguments": ["--vm-bytes", "60M", "--vm", "1"]
}
Example configration for influxdb:
{
"id": "influxdb",
"dockerimage": "influxdb",
"mem": "100",
"cpus": "1",
"sla": "singleton",
"nodetype": "analytics",
"notmetered": true,
"volumes": [
{
"host_path":"/analytics/log/influxdb",
"container_path":"/log/",
"mode":"rw"
}
],
"ports": [
{
"host_port":"31400",
"container_port":"8083"
},
{
"host_port":"31410",
"container_port":"8086"
}
]
}
Field-definitions:
id The name of the task. Charmander will internally add a timestamp to the id to make it unique
dockerimage The name of the docker image to start
mem Memory-allocation requested
cpus Percentage of cpu requested
sla Current supported sla-types are "one-per-node" (see cAdvisor) and "singleton" (see influxdb)
nodetype Node-types can be either "analytics" (for analytics node) and "lab" (for lab nodes)
notmetered Defines if the metrics of a task gets metered/collected or not
reshuffeable Defines if a task will be restarted by the "reshuffle" command. Typically set for simulators
ports Container-Port-mapping: "host_port" and "container_port" define intern and external port, in "bridge" mode only
volumes Volume-mapping for the container. "host_path", "container_path", and "mode" define the mapping of volumes
networkmode Supported modes: "bridge", "host", and "none", see Docker documentation for further explanation
arguments Additional arguments that will get passed to the ENTRY_POINT of the docker image
Example: Start stress60mb
curl -s -X POST -T loadsimulator/stress/stress60mb.json -H "Accept: application/json" -H "Content-Type: application/json" 172.31.1.11:7075/client/task
Which returns a confirmation that the task request got accepted and is now waiting for a matching offer from Mesos to get started
{"code":202,"message":"Run task: stress60"}
List Task
curl 172.31.1.11:7075/client/task
Returns a detailed list of all the tasks (scheduled and running).
Kill Task
Example: Kill all tasks whose task id starts with stress
curl -X DELETE 172.31.1.11:7075/client/task/stress
This should return a list of all the tasks that are planned to be killed. Only running tasks in that list will get deleted.
{"code":200,"message":"Kill task: stress60-1421970700191214531"}
Reshuffle Tasks
After you have collected intelligence about the running simulators you typically want to verify the impact it has on
future distribution. Reshuffle will restart all the currently running simulators ("reshuffleable": true) based on the
collected task-intelligence.
curl 172.31.1.11:7075/client/task/reshuffle
.. response you get.
{"code":202,"message":"Reshuffle tasks"}