如何在不重新启动的情况下更改docker容器的源代码？

Question

I have an application where I am using two docker containers. Say container and . Now B is dependent upon A. So when I start container 'B', I do mention the ip address of container 'A' in the command line.

Now I have a code change in the container 'A'. If I have to restart the container 'A' after change the source code then after restarting A's ip address will change. For that reason container 'B' will lose connection to 'A' and I have to restart 'B' with new ip address of 'A'.

Is there any docker approach where I can reserve the ip address for container 'A' always? or can I make the code changes without restarting the container 'A'? Please suggest.

Answer 1

You would need a service discovery like consul , combined with registrator in order to record A (in consul) each time it starts.

Then B could query consul in order to get A up-to-date IP address.

See " Automatic Docker Service Announcement with Registrator "

If your containers are not managed by the same docker daemon (meaning you are in a multi-hosts multi-nodes environment), you would use docker swarm (still in conjunction with consul ):

• " Deploying Containers with Docker Swarm and Docker Networking "
• " Atlassian Orchestration with Docker: multi-host support for the win! "

Answer 2

Two ways

Custom IP's

You can customise Docker networking inside a container far beyond what Docker allows on the command line with pipework by Jérôme Petazzoni .

It allows you to set the networking you would like so you could create a bridge, lets be inventive and call it br1 , for your containers and set specific IP's on each container or even use a dhcp server to allocate the IP's. This is a static config:

Start the aa container and add an IP on our bridge, br1 as eth0

→ docker run -d --net=none --name aa busybox cat
→ pipework br1 -eth0 aa 192.168.17.4/24

Start the bb container and add 192.168.17.5 on br1

→ docker run -d --net=none --name bb busybox cat
→ pipework br1 -eth0 bb 192.168.17.5/24

Now we know IP address of a container all the time

→ docker exec bb ping 192.168.17.4

Docker hosts file

Docker 1.8 enabled /etc/hosts container name sharing by default.

Docker 1.9 disabled this on the default network/bridge docker0 but will share container name information on a custom network

Create the network

→ docker create network shared
ce8fb4c6d6d379e87314952f4dfbcdcd6cf6c03416431721bb8f6001492575ea

Run a container named aa

→ docker run --name aa --net=shared -ti -d busybox sleep 5
754c7716cee65526cb2284066ccf2e495d494f5c351e281a89672aa5bdf1d84d

Ping by container name aa

→ docker run --name bb --net=shared -ti --rm busybox ping aa
PING aa (172.18.0.2): 56 data bytes
64 bytes from 172.18.0.2: seq=0 ttl=64 time=1.354 ms
64 bytes from 172.18.0.2: seq=1 ttl=64 time=0.081 ms
64 bytes from 172.18.0.2: seq=2 ttl=64 time=0.090 ms
64 bytes from 172.18.0.2: seq=3 ttl=64 time=0.114 ms

--- aa ping statistics ---
8 packets transmitted, 4 packets received, 50% packet loss
round-trip min/avg/max = 0.081/0.409/1.354 ms

Now the container has stopped after 5 seconds

→ docker run --name bb --net=shared -ti --rm busybox ping aa
ping: bad address 'aa'

Docker multi node hosts file

Docker natively supports multiple host networking via its overlay network driver . This does require an external service discovery agent to provide the kay/value store service but it means you can do this over multiple docker nodes too.

Answer 3

you do this with the --volumes-from flag. It would look like this:

$ docker run a --name a
$ docker run b --name b --volumes-from a

Docker will transparently handle the networking for you.

https://docs.docker.com/engine/userguide/dockervolumes/