I try to start a docker container with docker-py (Version 1.3.1). I want to map the container internal ports to different ports but fail to expose them properly.
我尝试使用docker-py(版本1.3.1)启动docker容器。我希望将容器内部端口映射到不同的端口,但不能正确地显示它们。
I do this like so:
我这样做:
def start_container(client, host_config, image_tagged_name, command):
print ("create_host_config", host_config.binds, host_config.port_bindings)
the_host_config = create_host_config(binds = host_config.binds,
port_bindings = host_config.port_bindings);
the_ports = host_config.port_bindings.values();
print ("create_container", image_tagged_name, command, the_ports, the_host_config)
cont_id = client.create_container(image=image_tagged_name, command=command, ports=the_ports, host_config=the_host_config)["Id"]
In the case at hand the output is as follows:
在手头的情况下,输出如下:
create_host_config ['/dbfiles/test:/opt/db'] {3001: 3000, 2425: 2424, 2481: 2480}
create_container test:test ./initdb.sh [3000, 2424, 2480] {'Binds': ['/dbfiles/test:/opt/db'], 'PortBindings': {'3001/tcp': [{'HostPort': '3000', 'HostIp': ''}], '2425/tcp': [{'HostPort': '2424', 'HostIp': ''}], '2481/tcp': [{'HostPort': '2480', 'HostIp': ''}]}}
docker ps tells me:
码头工人ps告诉我:
$ docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
169ad3ae0f63 test:test "./initdb.sh" 5 minutes ago Up 5 minutes 2424/tcp, 2480/tcp, 3000/tcp silly_pasteur
However if I give it mappings 3000 -> 3000, 2424 -> 2424 and 2480 -> 2480 it gives
但是如果我给它映射3000 -> 3000,2424 -> 2424和2480 -> 2480。
$ docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
cba483673bdd test:test "./initdb.sh" 53 minutes ago Up 5 minutes 0.0.0.0:2424->2424/tcp, 0.0.0.0:2480->2480/tcp, 0.0.0.0:3000->3000/tcp stupefied_ptolemy
The point is that from the commandline I can start the container with proper port mappings. That is
关键是,从命令行开始,我可以使用适当的端口映射启动容器。这是
docker run -d -p 3001:3000 -p 2425:2424 -p 2481:2480 -v /dbfiles/test:/opt/db localhost:5000/test:test /initdb.sh
mso - bidi - font - family: " times new roman "; mso - bidi - font - family: " times new roman "; mso - bidi - theme - font: minor - bidi ' > < / span > < span
gives the desired result.
给出了期望的结果。
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
7c1580e0ace9 localhost:5000/test:test "/initdb.sh" 8 seconds ago Up 6 seconds 0.0.0.0:2425->2424/tcp, 0.0.0.0:2481->2480/tcp, 0.0.0.0:3001->3000/tcp backstabbing_brahmagupta
However with docker-py I just can not figure out how to map the ports to different port numbers. What am I missing?
但是,对于docker-py,我只是不知道如何将端口映射到不同的端口号。我缺少什么?
2 个解决方案
#1
2
The issue was that docker-py puts the container ports first in its host configuration while the docker client puts them second. More interesting though is how I finally found it out. The trick was to install socat and then
问题是docker-py将容器端口放在主机配置中,而docker客户机将它们放在第二位。更有趣的是,我最终发现了它。关键是要安装socat。
$ socat -v UNIX-LISTEN:/tmp/debug, fork UNIX-CONNECT:/var/run/docker.sock
$ export DOCKER_HOST=unix:///tmp/debug
This allows to conveniently look into the traffic of the docker client as well as the docker-py client.
这样可以方便地查看docker客户端和docker-py客户机的流量。
I searched inside for the PortBindings strings. For the original client this gave me:
我在内部搜索PortBindings字符串。对于最初的客户,这给了我:
"PortBindings": {
"2424/tcp": [{"HostIp":"","HostPort":"2425"}],
"2480/tcp": [{"HostIp":"","HostPort":"2481"}],
"3000/tcp": [{"HostIp":"","HostPort":"3001"}]
}
While for my code it gave me
我的代码给了我。
"PortBindings": {
"2425/tcp": [{"HostPort": "2424", "HostIp": ""}],
"2481/tcp": [{"HostPort": "2480", "HostIp": ""}],
"3001/tcp": [{"HostPort": "3000", "HostIp": ""}]
},
This made everything obvious. The issue was not failure to expose the ports but wrong ordering of the ports.
这使得一切都显而易见。这个问题不是没有暴露港口,而是错误的港口秩序。
#2
1
You have to publish and expose the ports when using docker-py. (When you publish with docker run, the ports are implicitly exposed)
在使用docker-py时,必须发布和公开端口。(当您使用docker运行发布时,端口隐式暴露)
example:
例子:
container = config['connection'].create_container(
image=imageName,
name=containerName,
ports=[2424],
host_config=create_host_config(port_bindings={2424:2425})
)
#1
2
The issue was that docker-py puts the container ports first in its host configuration while the docker client puts them second. More interesting though is how I finally found it out. The trick was to install socat and then
问题是docker-py将容器端口放在主机配置中,而docker客户机将它们放在第二位。更有趣的是,我最终发现了它。关键是要安装socat。
$ socat -v UNIX-LISTEN:/tmp/debug, fork UNIX-CONNECT:/var/run/docker.sock
$ export DOCKER_HOST=unix:///tmp/debug
This allows to conveniently look into the traffic of the docker client as well as the docker-py client.
这样可以方便地查看docker客户端和docker-py客户机的流量。
I searched inside for the PortBindings strings. For the original client this gave me:
我在内部搜索PortBindings字符串。对于最初的客户,这给了我:
"PortBindings": {
"2424/tcp": [{"HostIp":"","HostPort":"2425"}],
"2480/tcp": [{"HostIp":"","HostPort":"2481"}],
"3000/tcp": [{"HostIp":"","HostPort":"3001"}]
}
While for my code it gave me
我的代码给了我。
"PortBindings": {
"2425/tcp": [{"HostPort": "2424", "HostIp": ""}],
"2481/tcp": [{"HostPort": "2480", "HostIp": ""}],
"3001/tcp": [{"HostPort": "3000", "HostIp": ""}]
},
This made everything obvious. The issue was not failure to expose the ports but wrong ordering of the ports.
这使得一切都显而易见。这个问题不是没有暴露港口,而是错误的港口秩序。
#2
1
You have to publish and expose the ports when using docker-py. (When you publish with docker run, the ports are implicitly exposed)
在使用docker-py时,必须发布和公开端口。(当您使用docker运行发布时,端口隐式暴露)
example:
例子:
container = config['connection'].create_container(
image=imageName,
name=containerName,
ports=[2424],
host_config=create_host_config(port_bindings={2424:2425})
)