默认情况下playbook中的任务执行时会一直保持连接,直到该任务在每个节点都执行完毕.有时这是不必要的,比如有些操作运行时间比SSH超时时间还要长.
解决该问题最简单的方式是异步执行它们,然后轮询直到任务执行完毕.
你也可以对执行时间非常长（有可能遭遇超时）的操作使用异步模式.
指定任务为异步执行
为了异步启动一个任务,指定其async最大超时时间以及轮询其状态的频率.如果你没有为 poll 指定值,那么默认的轮询频率是15秒钟。pool设置为0时，任务会立即返回，而不等待命令执行的结果，继续执行下面的任务。
例如下面的palybook
# asynctest.yml
---
- hosts: node1
tasks:
- shell: sleep 100 && hostname
async: 100
poll: 0
register: result
- debug: var=result
- async_status: jid={{ result.ansible_job_id }}
register: job_result
until: job_result.finished
retries: 30
第一个任务，指定shell任务为异步执行，100秒后任务超时失败。
第二个任务，获取异步任务的返回值，其目的是获取jid。
第三个任务，检查jid异步任务的状态，当异步任务的finished不为0时，异步任务执行成功。检查次数为30次，间隔5秒。
使用ansible-playbook执行任务
# ansible-playbook asynctest.yml -vv
Using /etc/ansible/ansible.cfg as config file
PLAYBOOK: asynctest.yml ********************************************************************************************************
1 plays in asynctest.yml
PLAY [node1] *******************************************************************************************************************
TASK [Gathering Facts] *********************************************************************************************************
ok: [192.168.77.129]
META: ran handlers
TASK [command] *****************************************************************************************************************
task path: /etc/ansible/asynctest.yml:5
changed: [192.168.77.129] => {"ansible_job_id": "402058509351.34914", "changed": true, "finished": 0, "results_file": "/root/.ansible_async/402058509351.34914", "started": 1}
TASK [debug] *******************************************************************************************************************
task path: /etc/ansible/asynctest.yml:10
ok: [192.168.77.129] => {
"changed": false,
"result": {
"ansible_job_id": "402058509351.34914",
"changed": true,
"finished": 0,
"results_file": "/root/.ansible_async/402058509351.34914",
"started": 1
}
}
TASK [async_status] ************************************************************************************************************
task path: /etc/ansible/asynctest.yml:12
FAILED - RETRYING: async_status (30 retries left).
FAILED - RETRYING: async_status (29 retries left).
FAILED - RETRYING: async_status (28 retries left).
FAILED - RETRYING: async_status (27 retries left).
FAILED - RETRYING: async_status (26 retries left).
FAILED - RETRYING: async_status (25 retries left).
FAILED - RETRYING: async_status (24 retries left).
FAILED - RETRYING: async_status (23 retries left).
FAILED - RETRYING: async_status (22 retries left).
FAILED - RETRYING: async_status (21 retries left).
FAILED - RETRYING: async_status (20 retries left).
FAILED - RETRYING: async_status (19 retries left).
FAILED - RETRYING: async_status (18 retries left).
FAILED - RETRYING: async_status (17 retries left).
FAILED - RETRYING: async_status (16 retries left).
FAILED - RETRYING: async_status (15 retries left).
FAILED - RETRYING: async_status (14 retries left).
FAILED - RETRYING: async_status (13 retries left).
changed: [192.168.77.129] => {"ansible_job_id": "402058509351.34914", "attempts": 19, "changed": true, "cmd": "sleep 100 && hostname", "delta": "0:01:40.005387", "end": "2017-06-21 17:07:31.015694", "finished": 1, "rc": 0, "start": "2017-06-21 17:05:51.010307", "stderr": "", "stderr_lines": [], "stdout": "node1", "stdout_lines": ["node1"]}
META: ran handlers
META: ran handlers
PLAY RECAP *********************************************************************************************************************
192.168.77.129 : ok=4 changed=2 unreachable=0 failed=0
使用ansible来执行异步
[root@master ansible]# ansible node1 -B 3600 -P 0 -m yum -a "name=ansible" -vv
Using /etc/ansible/ansible.cfg as config file
META: ran handlers
192.168.77.129 | SUCCESS => {
"ansible_job_id": "23974611070.37468",
"changed": true,
"finished": 0,
"results_file": "/root/.ansible_async/23974611070.37468",
"started": 1
}
META: ran handlers
META: ran handlers
参数说明
-B 3600 启用异步，超时时间3600,
-P 0 轮询时间为0
使用async_status来获取异步状态信息
[root@master ansible]# ansible node1 -m async_status -a "jid=23974611070.37468"
192.168.77.129 | SUCCESS => {
"ansible_job_id": "23974611070.37468",
"changed": false,
"finished": 1,
"msg": "",
"rc": 0,
"results": [
"ansible-2.3.1.0-1.el6.noarch providing ansible is already installed"
]
}
注意：在使用
command,win_command,shell,win_shell,raw模块时，默认是不会返回信息的，可以使用-o显示。
异步任务的状态文件
异步任务的状态文件以jid命名的方式存放在远端主机的用户目录下的.ansible_async目录，本次使用的是root连接远端，所以目录是/root/.ansible_async。
查看状态文件
[root@node1 ~]# tail /root/.ansible_async/402058509351.34914
{"changed": true, "end": "2017-06-21 17:07:31.015694", "stdout": "node1", "cmd": "sleep 100 && hostname", "start": "2017-06-21 17:05:51.010307", "delta": "0:01:40.005387", "stderr": "", "rc": 0, "invocation": {"module_args": {"creates": null, "executable": null, "_uses_shell": true, "_raw_params": "sleep 100 && hostname", "removes": null, "warn": true, "chdir": null}}, "warnings": []}
异步任务的返回值
{
"changed": false,
"result": {
"ansible_job_id": "402058509351.34914",
"changed": true,
"finished": 0,
"results_file": "/root/.ansible_async/402058509351.34914",
"started": 1
}
}
异步任务状态的返回值
可通过 async_status 获取，async_status 是读取远端的异步任务状态文件来获得任务状态，如果async设置的时间太短，可能会导致获取不到状态文件，因为还没生成这个文件。
例： ansible node1 -m async_status -a "jid=347674660626.31272"
执行中的返回值
{
"ansible_job_id": "801846426045.31757",
"changed": false,
"finished": 0,
"started": 1
}
执行后的返回值
{
"ansible_job_id": "402058509351.34914",
"attempts": 19,
"changed": true,
"cmd": "sleep 100 && hostname",
"delta": "0:01:40.005387",
"end": "2017-06-21 17:07:31.015694",
"finished": 1,
"rc": 0,
"start": "2017-06-21 17:05:51.010307",
"stderr": "",
"stderr_lines": [],
"stdout": "node1",
"stdout_lines": [
"node1"
]
}