|
MESOS 提供了Scheduler HTTP RESTful API. 不论是软件开发人员还是软件测试人员都可以通过这些API 进一步了解mesos的工作机制。本文讲述了如何用curl来访问这些API。
环境配置
1.mesos master，master的command options 设置如下：
MESOS_MASTER_OPTS="--work_dir=/var/lib/mesos --log_dir=/var/log/mesos" |
2 mesos slaves，每个slave的配置是一样的，2 CPU，996M memory。slave的command options设置如下
MESOS_SLAVE_OPTS="--master=$MESOS_MASTER_IP:5050 --log_dir=/var/log/mesos" |
注册Framework
1.注册framework是framework与mesos master通信的第一步。Framework注册成功之后，才会建立起和mesos master之间的联系。运行以下命令来注册一个framework。
# curl –vv --no-buffer –X POST -H "Content-Type:
application/json" -d@register.json
http://$MESOS_MASTER_IP:5050/master/api/v1/scheduler |
curl 后面的-d 指的是传送给API的data. 这个data是个json串。在命令行中写长的json串是个比较痛苦的事情，最好将json串写进一个文件。本例中该文件的名字为register.json. json文件怎么定义可以查看mesos.proto， scheduler.proto.
Response body like this:
* upload completely sent off: 257 out of 257 bytes
< HTTP/1.1 200 OK
< Date: Fri, 15 Apr 2016 05:56:03 GMT
< Mesos-Stream-Id:
b3a3e239-f0ad-43d3-b635-1c1d9c66566b
< Content-Type: application/json
< X-Cache: MISS from db03b04
< X-Cache-Lookup: MISS from db03b04:3128
< Transfer-Encoding: chunked
< Via: 1.1 db03b04 (squid/3.3.8)
< Connection: keep-alive < 70
{"subscribed":{"framework_id":{"value":"test13"}},
"type":"SUBSCRIBED"}20
{"offers":{"offers":[{"agent_id":
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S1"},
"framework_id":{"value":"test13"},
"hostname":"slave1","id":
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O4"},
"resources":[{"name":"cpus","role":"*","scalar":
{"value":2.0},"type":"SCALAR"},
{"name":"mem","role":"*","scalar":
{"value":996.0},"type":"SCALAR"},
{"name":"disk","role":"*","scalar":
{"value":30509.0},"type":"SCALAR"},
{"name":"ports","ranges":{"range":
[{"begin":31000,"end":32000}]},
"role":"*","type":"RANGES"}],"url":
{"address":{"hostname":"slave1",
"ip":"9.111.254.36","port":5051},
"path":"\/slave(1)","scheme":"http"}},
{"agent_id":{"value":"
b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S0"},
"framework_id":{"value":"test13"},"hostname":"slave0",
"id":{"value":"
b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O5"},
"resources":[{"name":"cpus","role":"*",
"scalar":{"value":2.0},"type":"SCALAR"},
{"name":"mem","role":"*","scalar":{"value":996.0},
"type":"SCALAR"},{"name":"disk","role":"*",
"scalar":{"value":30509.0},"type":"SCALAR"},
{"name":"ports","ranges":{"range":
[{"begin":31000,"end":32000}]},
"role":"*","type":"RANGES"}],
"url":{"address":{"hostname":"slave0",
"ip":"9.111.254.62","port":5051},
"path":"\/slave(1)","scheme":"http"}
}]},"type":"OFFERS"} |
Mesos-Stream-Id, 作为该framework的唯一标识。在后面的actions 像launch task， decline offers等都需要用到它.
framework 注册成功之后，如果master 有资源的话，该framework会收到master发送来的offers。
register.json文件内容示例：
{ "framework_id": {"value" : "test13"},
"type":"SUBSCRIBE",
"subscribe":{
"framework_info":{ "user":"root", "name":"test13",
"failover_timeout":60,
"role":"aa", "id":{"value":"test13"},
"principal":"test",
"capabilities":{"type":"REVOCABLE_RESOURCES"} },
"force":true } } |
2.使用已注册好的framework 来launch task。
# curl -vv --no-buffer -X POST -H "Content-Type:
application/json" -H
"Mesos-Stream-Id:b3a3e239-f0ad-43d3-b635-1c1d9c66566b"
-d@launch_task.json
http://$MESOS_MASTER_IP:5050/master/api/v1/scheduler |
Response body like this:
* Done waiting for 100-continue
< HTTP/1.1 202 Accepted
< Date: Wed, 13 Apr 2016 05:37:44 GMT
< Content-Length: 0
< X-Cache: MISS from db03b04
< X-Cache-Lookup: MISS from db03b04:3128
< Via: 1.1 db03b04 (squid/3.3.8)
< Connection: keep-alive |
launch_task.json文件内容示例：
{
"framework_id":{"value":"test13"},
"type":"ACCEPT",
"accept":{
"offer_ids":[
{"value":"81c111bd-f27d-41a4-b184-64090dec3048-O428"}
],
"filters":{},
"operations":[
{
"type":"LAUNCH",
"launch":{"task_infos":[{
"name":"task 00012",
"task_id":{"value": "00012"},
"agent_id":{"value": "5bb4aba0-628b-4309-b36f-767db1ebb7f4-S0"},
"resources":[
{
"name":"mem",
"type":"SCALAR",
"scalar":{"value":996.0}
},
{
"name":"cpus",
"type":"SCALAR",
"scalar":{"value":1.0}
}
],
"command":{"value": "/bin/sleep 3000s"}
}
]
}
}
]
}
} |
Result
查看framework output ， framework 收到了UPDATE event， 这是MESOS master发给framework更新task 状态的UPDATE
{"type":"UPDATE","update":
{"status":{"agent_id":
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S1"},
"container_status":{"network_infos":
[{"ip_address":"9.111.254.36",
"ip_addresses":[{"ip_address":"9.111.254.36"}]}]},
"executor_id":{"value":"00016"},
"source":"SOURCE_EXECUTOR",
"state":"TASK_RUNNING",
"task_id":{"value":"00016"},
"timestamp":1460526113.05411,
"uuid":"H9KoCldtRTu0T7halLxs9w=="}}}20
{"type":"HEARTBEAT"}391 |
注意：master 会将当前offer剩余的资源作为新的offer 提供给其它的framework。形成一个新的资源。
3. Framework 收到task status 之后，需要发acknowledge event 给master ，表示已经收到。master就会一直不停的发同样的event给framework直到它收到该event的acknowledge
# curl -vv --no-buffer -X POST -H "Content-Type:
application/json"
-H "Mesos-Stream-Id:b3a3e239-f0ad-43d3-b635-1c1d9c66566b"
-d@acknowledge.json
http://$MESOS_MASTER_IP:5050/master/api/v1/scheduler |
Response body like this:
* upload completely sent off:
264 out of 264 bytes
< HTTP/1.1 202 Accepted
< Date: Wed, 13 Apr 2016 06:01:31 GMT
< Content-Length: 0
< X-Cache: MISS from db03b04
< X-Cache-Lookup: MISS from db03b04:3128
< Via: 1.1 db03b04 (squid/3.3.8)
< Connection: keep-alive |
acknowledge.json的文件内容示例如下：
{ "framework_id" :
{"value" : "test13"},
"type" : "ACKNOWLEDGE",
"acknowledge" :
{ "agent_id" :
{"value" :
"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S1"},
"task_id" :
{"value" : "00016"},
"uuid" :
"H9KoCldtRTu0T7halLxs9w==" } } |
Result
在Framework的output里，同样UUID的event信息不再出现
4. 当task 运行完毕的时候，它所占用的资源会被释放掉，从而一个新的资源产生了，master又将该资源按照DRF算法发给了相应的framework。
5. Decline offer, 如果目前分配给framework的offer不合适，framework可以decline it。
# curl -vv --no-buffer -X POST -H "Content-Type:
application/json"
-H "Mesos-Stream-Id:b3a3e239-f0ad-43d3-b635-1c1d9c66566b"
-d@decline.json
http://$MESOS_MASTER_IP:5050/master/api/v1/scheduler |
Response body like this
* upload completely sent off: 214 out of 214 bytes
< HTTP/1.1 202 Accepted
< Date: Wed, 13 Apr 2016 06:17:56 GMT
< Content-Length: 0
< X-Cache: MISS from db03b04
< X-Cache-Lookup: MISS from db03b04:3128
< Via: 1.1 db03b04 (squid/3.3.8)
< Connection: keep-alive
< |
decline.json 文件内容示例：
{ "framework_id" :
{"value" : "test13"},
"type" :
"DECLINE", "decline" :
{ "offer_ids" : [
{"value" : "b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O15"}
] } } |
Result
被decline的offer不再可用。
6. 如果在mesos cluster 运行很多个task，在mesos cluster 系统里面就会不停的创造出很多的资源碎片，这些碎片都不能单独的launch task的时候，该怎么办呢？有两个办法可以合并同一个slave上面的资源碎片。
6.1 ACCEPT event
在launch task的时候，将slave上面的碎片资源写成list，这样会自动合并碎片资源。注意这些资源碎片一定是属于同一个slave
Launch_task.json 文件内容示例：
{
"framework_id":{"value":"test13"},
"type":"ACCEPT",
"accept":{
"offer_ids":[
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O14"},
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O16"}
],
"filters":{},
"operations":[
{
"type":"LAUNCH",
"launch":{"task_infos":[{
"name":"task 00017",
"task_id":{"value": "00017"},
"agent_id":{"value": "b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S1"},
"resources":[
{
"name":"mem",
"type":"SCALAR",
"scalar":{"value":996.0}
},
{
"name":"cpus",
"type":"SCALAR",
"scalar":{"value":2.0}
}
],
"command":{"value": "/bin/sleep 180s"}
}
]
}
}
]
} } |
当task运行完毕，mesos master就会将这些资源作为整块资源按照DRF算法发给相应的framework
6.2 decline event 合并碎片资源
将各个碎片资源decline掉， 这样也能整合碎片资源。 例如有三个offers，这3个offer都属于同一个slave。decline 这些offer之后，mesos master会自动整合这些碎片资源，然后按照DRF算法将合并后的资源发给相应的framework
{"offers":{"offers":[{"agent_id":
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S1"},
"framework_id":{"value":"test13"}
,"hostname":"slave1","id":
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O24"},
"resources":[{"name":"cpus","role":"*",
"scalar":{"value":1.0},"type":"SCALAR"},
{"name":"mem","role":"*","scalar":{"value":512.0},
"type":"SCALAR"}],"url":{"address":
{"hostname":"slave1","ip":"9.111.254.36","port":5051},
"path":"\/slave(1)","scheme":"http"}}]},"type":"OFFERS"}
{"offers":{"offers":[{"agent_id":
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S1"},
"framework_id":{"value":"test13"},"hostname":
"slave1","id":{"value":
"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O26"},
"resources":[{"name":"cpus","role":"*",
"scalar":{"value":1.0},"type":"SCALAR"},
{"name":"mem","role":"*","scalar":
{"value":256.0},"type":"SCALAR"}],
"url":{"address":{"hostname":"slave1",
"ip":"9.111.254.36","port":5051},"path":
"\/slave(1)","scheme":"http"}}]},"type":"OFFERS"}
{"offers":{"offers":[{"agent_id":{"value":
"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S1"},
"framework_id":{"value":"test13"},"hostname":
"slave1","id":{"value":
"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O27"}
,"resources":[{"name":"ports","ranges":
{"range":[{"begin":31000,"end":32000}]},
"role":"*","type":"RANGES"},
{"name":"mem","role":"*","scalar":
{"value":228.0},"type":"SCALAR"},
{"name":"disk","role":"*",
"scalar":{"value":30509.0},
"type":"SCALAR"}],
"url":{"address":
{"hostname":"slave1",
"ip":"9.111.254.36"
,"port":5051},"path":
"\/slave(1)","scheme":"http"}
}]},"type":"OFFERS"} |
decline offer 内容如下：
{
"framework_id"
: {"value" : "test13"},
"type"
: "DECLINE",
"decline"
: {
"offer_ids" : [
{"value" : "b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O24"},
{"value" : "b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O26"},
{"value" : "b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-O27"}
] } }
|
decline 成功以后，一个整合过的offer就会发送给相应的framework
7. kill task。
kill event主要用于kill special task，一次kill一个task
# curl -vv --no-buffer -X POST -H "Content-Type:
application/json" -H
"Mesos-Stream-Id
:b3a3e239-f0ad-43d3-b635-1c1d9c66566b"
-d@kill_task.json
http://$MESOS_MASTER_IP:5050/master/api/v1/scheduler |
Response body like this
* upload completely sent off: 211 out of 211 bytes
< HTTP/1.1 202 Accepted
< Date: Fri, 15 Apr 2016 13:08:25 GMT
< Content-Length: 0
< X-Cache: MISS from db03b04
< X-Cache-Lookup: MISS from db03b04:3128
< Via: 1.1 db03b04 (squid/3.3.8)
< Connection: keep-alive |
kill_task.json 文件内容示例：
{ "framework_id"
: {"value" : "test13"},
"type" :
"KILL", "kill"
: { "task_id" :
{"value" : "00020"},
"agent_id" :
{"value" : "b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S1"}
} } |
Result
The output in framework like this:
{"type":"UPDATE","update":
{"status":{"agent_id":
{"value":"b5e0c23f-237e-4398-816f-ad1a4aa7d3f3-S0"},
"container_status":
{"network_infos":
[{"ip_address":"9.111.254.62",
"ip_addresses":[{"ip_address":"9.111.254.62"}]}]},
"executor_id":{"value":"00012"},
"message":"Command terminated with signal Terminated",
"source":"SOURCE_EXECUTOR","state":
"TASK_KILLED","task_id":{"value":"00012"}
,"timestamp":1460725705.73596,"uuid":
"mKSkxCycRLKS8y4YA2HM4A=="}}} |
该task的资源被释放重新利用。如果只有一个framework的话，在当前framework的output中就会有一个新的offer。
8. TEARDOWN framework
# curl -vv --no-buffer -X POST -H "Content-Type:
application/json" -H
"Mesos-Stream-Id:
b3a3e239-f0ad-43d3-b635-1c1d9c66566b"
-d@teardown.json
http://9.111.254.199:5050/master/api/v1/scheduler |
Response body like this
* upload completely sent off:
77 out of 77 bytes
< HTTP/1.1 202 Accepted
< Date: Fri, 15 Apr 2016 13:25:49 GMT
< Content-Length: 0
< X-Cache: MISS from db03b04
< X-Cache-Lookup: MISS from db03b04:3128
< Via: 1.1 db03b04 (squid/3.3.8)
< Connection: keep-alive |
teardown.json 文件内容如下：
{ "framework_id"
: {"value" : "test13"},
"type"
: "TEARDOWN" } |
Result
注册framework的进程会自动退出。Framework 变成inactive
More information about mesos RESTful api ，please refer to
http://mesos.apache.org/documentation/latest/scheduler-http-api/
|
