跨docker容器共享预编译资产-Sharingprecompiledassetsacrossdockercontainers

17  

I'm having the same issue. Here's what I'm currently working on:

我有同样的问题。这是我目前正在做的事情：

Option 1: use a single image for both nginx and the app

This way, I can build the image once (with the app, precompiled assets and nginx), then run two instances of it: one running the app server, and another for the nginx frontend:

这样，我可以构建一次图像（使用app，预编译资产和nginx），然后运行它的两个实例：一个运行app服务器，另一个运行nginx前端：

docker build -t hello .
docker run --name hello-app hello rackup
docker run --name hello-web -p 80:80 --link hello-app:app hello nginx

Not pretty, but very easy to set up and upgrade.

不漂亮，但很容易设置和升级。

Option 2: use a shared volume, and precompile assets as a job

Shared volumes cannot be updated in the build process, but can be updated by a container instance. So we can run our rake task to precompile the assets just before running our app:

共享卷无法在构建过程中更新，但可以由容器实例更新。因此，我们可以在运行应用程序之前运行我们的rake任务来预编译资产：

docker build -t hello .
docker run -v /apps/hello/assets:/app/public/assets hello rake assets:precompile
docker run --name hello-app hello rackup
docker run --name hello-web -p 80:80 --link hello-app:app -v /apps/hello/assets:/usr/share/nginx/html/assets nginx

This looks like a more robust option, but will require a more complex instrumentation. I'm leaning towards this option, however, since we'll need a separate job for database migrations anyway.

这看起来像一个更强大的选项，但需要更复杂的仪器。但是，我倾向于这个选项，因为无论如何我们都需要一个单独的数据库迁移工作。

Option 3: distribute the assets to a CDN on build-time

Your Dockerfile can upload the resulting assets directly to a CDN. Then you configure your Rails app to use it as the asset_host. Something like:

您的Dockerfile可以将生成的资产直接上传到CDN。然后配置Rails应用程序以将其用作asset_host。就像是：

RUN rake assets:precompile && aws s3 sync public/assets s3://test-assets/

I'm currently experimenting with this option. Since I'm using Amazon CloudFront, looks like I can just sync the resulting assets to S3 using the AWS CLI. There's also a gem for that (asset_sync), but it looks stale.

我正在尝试这个选项。由于我使用的是Amazon CloudFront，看起来我可以使用AWS CLI将生成的资产同步到S3。还有一个宝石（asset_sync），但它看起来很陈旧。

The downside is that you'll have to send the needed AWS credentials to the build context or the Dockerfile itself – this might need committing them to your source repository, if you're using an automated build.

缺点是您必须将所需的AWS凭证发送到构建上下文或Dockerfile本身 - 如果您使用的是自动构建，则可能需要将它们提交到源存储库。

5  

I recommend to precompile assets inside of Docker container because in this case you can without any problems run it to any environment otherwise before run docker container you should copy shared precompilled assets to new environment, mount them and only after it run docker container.

我建议在Docker容器内部预编译资产，因为在这种情况下，您可以毫无问题地将其运行到任何环境，否则在运行docker容器之前，您应该将共享的预编译资产复制到新环境，安装它们并且仅在它运行docker容器之后。

The other problem related to version of assets. I mean that your code must use assets compatible with it otherwise you will have different problems with CSS and JS. So, this is the second point why you should precompile assets inside of Docker container.

另一个问题与资产版本有关。我的意思是你的代码必须使用与之兼容的资产，否则你将遇到CSS和JS的不同问题。所以，这是你应该在Docker容器中预编译资产的第二点。

0  

For production, @maxd is right: you should add the compiled assets to your images. However, for development I think you are right. Use a volume container that is used by:

对于生产，@ max是正确的：您应该将编译后的资产添加到图像中。但是，对于开发，我认为你是对的。使用以下使用的卷容器：

• the rails container to store the precompiled assets
• rails容器用于存储预编译资产
• the nginx container to serve the assets
• 用于服务资产的nginx容器

An alternative would be mounting a host directory into the containers.

另一种方法是将主机目录安装到容器中。

You can find more help in the Managing Data in Containers User Guide

您可以在“管理容器中的数据用户指南”中找到更多帮助

0  

Define a VOLUME in Dockerfile

在Dockerfile中定义一个VOLUME

And then use --volumes-from during run of another container

然后在另一个容器的运行期间使用--volumes-from