D.I.Y. mesos framework part 2
Last time we launched a simple Mesos framework. Now let’s see how to interact with it.
Prequisite
Must complete all the steps from Part One
Accept an offer and launch a task
The next step is to accept the offer from the Mesos Master and launch a custom task on our running Mesos agent. This brings us into contact with another part of the Mesos architecture: an Executor. An Executor is the process on the Mesos agent which is used to launch and monitor a task. While it is possible to write a custom Executor, for this example we’ll just be using the Mesos provided one.
Mesos provides a simple executor that can execute shell commands and Docker containers on behalf of the framework scheduler; enough functionality for a wide variety of framework requirements.
To use the default executor, we simply need to include a command as a top-level field under task_infos. If we wanted to use a custom executor, we would nest command inside an executor node.
Command below. Remember to update the values for:
mesos-stream-id
framework_id
offer_ids
agent_idIn terminal 4:
(due to the size of this outrageous JSON blob I recommend dropping this command into POSTMAN to avoid weird copy/paste issues in a terminal)
curl -X POST \
http://localhost:5050/api/v1/scheduler \
-H 'content-type: application/json' \
-H 'mesos-stream-id: 274bcc35-78db-42ac-8a10-b02944c5a6eb' \
-d '{
"framework_id": {
"value": "089a7212-4aba-4aac-8425-7337161ece30-0000"
},
"type": "ACCEPT",
"accept": {
"offer_ids": [
{
"value": "089a7212-4aba-4aac-8425-7337161ece30-O0"
}
],
"operations": [
{
"type": "LAUNCH",
"launch": {
"task_infos": [
{
"name": "My Task",
"task_id": {
"value": "shell hello world loop"
},
"agent_id": {
"value": "089a7212-4aba-4aac-8425-7337161ece30-S0"
},
"command": {
"shell": true,
"value": "echo starting; while [[ 1 -eq 1 ]]; do echo helloWorld; sleep 5; done"
},
"resources": [
{
"allocation_info": {
"role": "hello_world_role"
},
"name": "cpus",
"role": "*",
"type": "SCALAR",
"scalar": {
"value": 1
}
},
{
"allocation_info": {
"role": "hello_world_role"
},
"name": "mem",
"role": "*",
"type": "SCALAR",
"scalar": {
"value": 128
}
}
]
}
]
}
}
],
"filters": {
"refuse_seconds": 5
}
}
}'Mesos responds with a 202 indicating that it Accepted the request but has not processed it. This is a good example of the Actor model, one-way communication implemented by Mesos.
Flipping thru the open terminals here is some interesting output:
Master node:
Status update TASK_RUNNING (UUID: 3abe8974-9d80-4374-9025-aae5d37cda1f) for task shell hello world loop of framework
Forwarding status update TASK_RUNNING (UUID: 3abe8974-9d80-4374-9025-aae5d37cda1f)In our Framework terminal:
"type":"UPDATE" ... "state":"TASK_RUNNING" ...Back in the browser at http://localhost:5050/#/, we should see the shell hello world loop task under active tasks! Clicking into the sandbox, then the stdout will reveal our looping terminal output.
We’ve just deployed a task! Are you feeling the cloud??
Acknowledge a status update from Mesos
To be a good citizen we need to acknowledge the status update from Mesos. (the Framework terminal will indicate that )
In Terminal 4:
curl -X POST \
http://localhost:5050/api/v1/scheduler \
-H 'content-type: application/json' \
-H 'mesos-stream-id: 6c83f87e-e1d2-47c7-be1f-43f71d4794a6' \
-d '{
"framework_id": {
"value": "f6480abb-914d-47e4-a802-b123c7925e7a-0000"
},
"type": "ACKNOWLEDGE",
"acknowledge" : {
"agent_id" : {"value" : "f6480abb-914d-47e4-a802-b123c7925e7a-S0"},
"task_id" : {"value" : "shell hello world loop"},
"uuid" : "F+lGzCYnTam7KN0gWOq+6w=="
}
}'Flipping to the Master node terminal:
Processing ACKNOWLEDGE callNice.
Launch a Docker container
Now scroll back up a bit and you will find another Offer from the Mesos master. Let’s accept that offer and launch a Docker task!
curl -X POST \
http://localhost:5050/api/v1/scheduler \
-H 'content-type: application/json' \
-H 'mesos-stream-id: 274bcc35-78db-42ac-8a10-b02944c5a6eb' \
-d '{
"framework_id": {
"value": "089a7212-4aba-4aac-8425-7337161ece30-0000"
},
"type": "ACCEPT",
"accept": {
"offer_ids": [
{
"value": "089a7212-4aba-4aac-8425-7337161ece30-O1"
}
],
"operations": [
{
"type": "LAUNCH",
"launch": {
"task_infos": [
{
"name": "nginx",
"task_id": {
"value": "nginx"
},
"agent_id": {
"value": "089a7212-4aba-4aac-8425-7337161ece30-S0"
},
"command": {
"shell": true,
"value": "docker run --name mesos-nginx -p 8080:80 nginx"
},
"resources": [
{
"allocation_info": {
"role": "hello_world_role"
},
"name": "cpus",
"role": "*",
"type": "SCALAR",
"scalar": {
"value": 1
}
},
{
"allocation_info": {
"role": "hello_world_role"
},
"name": "mem",
"role": "*",
"type": "SCALAR",
"scalar": {
"value": 128
}
}
]
}
]
}
}
],
"filters": {
"refuse_seconds": 5
}
}
}'Now http://localhost:8080/ will show nginx running on our system! Sweet.
Decline an offer
back in Terminal 4:
curl -X POST \
http://localhost:5050/api/v1/scheduler \
-H 'content-type: application/json' \
-H 'mesos-stream-id: 274bcc35-78db-42ac-8a10-b02944c5a6eb' \
-d '{
"framework_id": {
"value": "089a7212-4aba-4aac-8425-7337161ece30-0000"
},
"type": "DECLINE",
"decline" : {
"offer_ids" : [
{"value" : "089a7212-4aba-4aac-8425-7337161ece30-O4"}
],
"filters" : {"refuse_seconds" : 5.0}
}
}'Kill a task
back in Terminal 4:
curl -X POST \
http://localhost:5050/api/v1/scheduler \
-H 'content-type: application/json' \
-H 'mesos-stream-id: 274bcc35-78db-42ac-8a10-b02944c5a6eb' \
-d '{
"framework_id": {
"value": "089a7212-4aba-4aac-8425-7337161ece30-0000"
},
"type": "KILL",
"kill" : {
"task_id" : {"value" : "nginx"},
"agent_id" : {"value" : "089a7212-4aba-4aac-8425-7337161ece30-S0"}
}
}'Processing KILL call for task 'nginx'
Telling agent 089a7212-4aba-4aac-8425-7337161ece30-S0 at slave(1)@192.168.1.12:5051 (192.168.1.12) to kill task nginx(note that the Task stays in “active” status in the UI, until we acknowledge the update from the Mesos master.)
–
fin