ab-nginx/README.md

ab-nginx

Containerized fully-functional implementation of NGINX running on Alpine as a fully NON-ROOT user. The container by default is a 'blank slate' that just serves files out of the box. Changing configuration, server blocks and content is accomplished with bind-mounts using a sensible, simple directory structure. The container auto-detects mounted certificates and switches to TLS automatically. Available Helper scripts make certificate mounting easier, allow for custom docker networks and more. The container by default can be used as a Let’s Encrypt endpoint with tools like certbot.

Version 5.x adds the following features:

• container auto-generates missing Diffie-Hellman Parameters (‘dhparams’) when using TLSv1.2
• included a script within the container to generate self-signed certificates with a group readable private key so they can be shared between services in a stack by aligning container and host GIDs
• helper scripts have been moved out of the ‘releases’ and hosted separately as straight-downloads so there is no need to clone the entire repo -- located at: https://asifbacchus.dev/public/docker/nginx/ab-nginx/

Contents

• Alternate repository
• Signed containers
• Documentation and scripts
• Permissions
  • Option 1: rebuild container with different UID/GID
  • Option 2: specify UID/GID at runtime
• Container layout
  • Content directory
  • Configuration directory
• Quick-start
  • Mounting content
  • Mounting configurations
  • Mounting server-blocks
• TLS
  • Generate a self-signed certificate
• Environment variables
• Shell mode
  • Drop to shell before NGINX loads
  • Enter a running container
• Logs
• Final thoughts

Alternate repository

Throughout this document, I reference my repository on DockerHub (asifbacchus/ab-nginx:tag). You may also feel free to pull directly from my private registry instead which is guaranteed to have the most up-to-date releases. Simply use docker.asifbacchus.dev/nginx/ab-nginx:tag.

Signed containers

Starting with the 5.x releases, I am no longer using Docker Notary to sign images. Instead, I’m using CodeNotary. This has several advantages, most notably that it doesn’t matter whether you use DockerHub or my private repo they will both have the same verifiable signature. To verify the signature, you would have to use CodeNotary’s vcn tool -- visit their site for the most up-to-date instructions.

Documentation and scripts

Check out the repo wiki for detailed examples and documentation about the container and the helper scripts.

Permissions

The container does NOT run under the root account. It runs under a user named www-docker with a UID and GID of 8080. This means any files you mount into the container need to be readable (and/or writable depending on your use-case) by UID 8080 or GID 8080. This does not just mean content files, it also includes configurations, server-blocks and certificates! Before mounting your files, ensure this is the case. There are more detailed instructions in the wiki if you need help setting file permissions.

This is a significant change versus most other NGINX implementations/containers where the main process is run as root and the worker processes run as a limited user. In those cases, permissions don’t matter since NGINX can always use the root account to read any files (and especially certificates!) it needs. Please understand this difference.

Most often you will end up wanting to change the container user’s GID so that you can assign it and related services appropriate permissions. To do this, you have two options:

Option 1: rebuild container with different UID/GID

If you are integrating this container as part of a stack, then it might just make sense to “hard-code” the UID and/or GID to whatever works best for your environment. Fortunately, that’s pretty easy. Clone the git repo and rebuild the container:

# clone the repo
cd /usr/local/src
git clone https://git.asifbacchus.dev/ab-docker/ab-nginx

# change directory and build
cd ab-nginx/build
docker build --build-arg UID=xxxx --build-arg GID=yyyy --build-arg BUILD_DATE=$(date -u +'%Y-%m-%dT%H:%M:%SZ') -t name:tag .

• UID=xxxx: optional -- replace ‘xxxx’ with desired UID for www-docker user, in most cases the default is fine
• GID=yyyy: replace ‘yyyy’ with desired GID for www-docker user --> this is probably what you want to change
• BUILD_DATE: optional -- applies container build date in a standardized label
• name:tag: you may, of course, name and tag your container anything you like

Option 2: specify UID/GID at runtime

You may find it easier and more flexible to just specify the UID and/or GID at runtime. Again, you most likely want to set the GID so that it matches, say, your webdev user group and the container can then read and serve those files. Pretty easy, just add --user "uid:gid" to your docker run command. Here’s an example where the GID is changed to 6000:

docker run -d ... --user "8080:6000" ... asifbacchus/ab-nginx:latest

Container layout

Content directory

All content is served from the NGINX default /usr/share/nginx/html directory within the container. The default set up serves everything found here and in all child directories. Bind-mount your content to the container’s webroot: -v /my/webstuff:/usr/share/nginx/html. Remember that the container UID or GID must be able to read these files!

Configuration directory

All configuration is in the /etc/nginx directory and its children. Here is the layout of that directory within the container:

/etc/nginx
├── config
│   └── **add configuration files here or replace the whole directory**
├── sites
│   ├── 05-nonsecured.conf
│   ├── 05-secured.conf.disabled
│   └── **add additional server-block files or replace whole directory**
├── ssl-config
│   ├── mozIntermediate_ssl.conf.disabled
│   └── mozModern_ssl.conf.disabled
	 └── (SSL configuration – container manages this)
├── errorpages.conf – (pre-configured fun error pages, you can override)
├── health.conf – (health-check endpoint, best to not touch this)
├── nginx.conf – **main NGINX configuration file, replace if really necessary**
├── server_names.conf – (list of hostnames, updated via environment variable)
├── ssl_certs.conf – (hard-coded for the container, best not to touch)

Locations with *starred descriptions* are designed to be overwritten via bind-mounts to customize the container. For more details on all of these files and what they do, please refer to the repo wiki. Remember that the container UID or GID needs to be able to read any files you choose to bind-mount over the container defaults!

Quick-start

At its most basic, all you need to do is mount a directory with content to serve. For more advanced deployments, you can also mount various configurations. In most cases, you’ll also want to mount certificates so that SSL/TLS is an option. Remember that the container’s UID/GID must be able to read everything you bind-mount! Let’s run through some examples:

Mounting content

Simply bind-mount whatever you want served to /usr/share/nginx/html:

docker run -d --name ab-nginx --restart unless-stopped \
  -p 80:80 \
  -v ~/web:/usr/share/nginx/html \
  asifbacchus/ab-nginx

Mounting configurations

Any .conf files found in /etc/nginx/config will be loaded after nginx.conf and thus, take precedence. All config files are read into the HTTP-context. Please note: only files ending in .conf will be read by the container!

I suggest dividing your configurations into various files organized by type (i.e. headers.conf, buffers.conf, timeouts.conf, etc.) and putting them all into one directory and bind-mounting that to the container:

docker run -d --name ab-nginx --restart unless-stopped \
  -p 80:80 \
  -v ~/web:/usr/share/nginx/html \
  -v ~/nginx/config:/etc/nginx/config:ro \
  asifbacchus/ab-nginx

If you need to change configuration settings, make the changes on the host and save the file(s). Then, restart the container to apply the change:

docker restart ab-nginx

If you want the container to ignore a specific set of configuration options, say you’re testing something, then rename the file with those configuration options using any extension other than .conf. I usually use .conf.disabled. Restart the container and that file will be ignored.

More details and examples are found in the wiki.

Mounting server-blocks

If you just want to serve static content from your content/webroot directory, then you can ignore this section entirely 😄.

Otherwise, any files found in the /etc/nginx/sites directory in the container will be loaded after the configuration files. These files are meant to define the SERVER-context. The container has both a secure and non-secure default server block that simply serves everything found in the webroot. Depending on your SSL configuration, the container enables the correct block. You can add additional server blocks or you can override these default servers entirely by bind-mounting over the directory:

# add another server block definition that listens on port 8080
docker run -d --name ab-nginx --restart unless-stopped \
  -p 80:80 \
  -p 8080:8080 \
  -v ~/web:/usr/share/nginx/html \
  -v ~/webapp.conf:/etc/nginx/sites/webapp.conf:ro \
  asifbacchus/ab-nginx

# override default server-blocks entirely (use your own)
docker run -d --name ab-nginx --restart unless-stopped \
  -p 80:80 \
  -v ~/web:/usr/share/nginx/html \
  -v ~/nginx/servers:/etc/nginx/sites:ro \
  asifbacchus/ab-nginx

More details and examples are found in the wiki.

TLS

The container will automatically update its configuration to use provided certificates. The examples below assume you have all required files in one directory, but you can also mount them all separately. The required files and their locations in the container are:

file type	container-location
Full-chain certificate (certificate concatenated with intermediates and root CA)	/certs/fullchain.pem
Private key	/certs/privkey.pem
Certificate chain (intermediates concatenated with root CA)	/certs/chain.pem
DH Parameters file (Container will generate this file if not provided) (NOT required for TLS 1.3-only mode)	/certs/dhparam.pem

Once those files are available, you can run the container as follows:

# TLS 1.2 (allows TLSv1.2 and TLSv1.3)
docker run -d --name nginx --restart unless-stopped \
  -p 80:80 \
  -p 443:443 \
  -v ~/web:/usr/share/nginx/html \
  -v ~/certs:/certs:ro \
  -e SERVER_NAMES="domain.tld www.domain.tld" \
  asifbacchus/ab-nginx:latest

# TLS 1.3 only mode
docker run -d --name nginx --restart unless-stopped \
  -p 80:80 \
  -p 443:443 \
  -v ~/web:/usr/share/nginx/html \
  -v ~/certs:/certs:ro \
  -e SERVER_NAMES="domain.tld www.domain.tld" \
  -e TLS13_ONLY=TRUE
  asifbacchus/ab-nginx:latest

The container will load a secure configuration automatically, require SSL connections and redirect HTTP to HTTPS. If you want to enforce HSTS, simply set the HSTS environment variable to true by adding -e HSTS=TRUE before specifying the container name. Careful about doing this while testing though! Also, certificates should always be mounted read-only (:ro) for security reasons!

You may have noticed I also specified the SERVER_NAMES variable. This is necessary or SSL will not work since the hostname the server responds to must match the certificate being presented. Make sure you set this environment variable to match your certificates!

N.B. If you are using your own server-blocks, then this environment variable is NOT required – it is only used by the container when auto-configuring the default server-blocks.

If you want to integrate with Let's Encrypt, please refer to the wiki.

Finally, I’d remind you once again that the container’s UID/GID must be able to read your certificate files! It is common practice to restrict the private key to root readability only (i.e. chown root:root & chmod 600/400) but, that would stop the NGINX user in the container from reading it and NGINX will exit with an error. I address ways to allow your certificate files to remain secure but still readable by the NGINX user in the wiki. As a quick hint, it’s easiest to accomplish by changing the container GID!

Generate a self-signed certificate

If you are testing a set-up or for whatever other reason want to use a self-signed certificate, the container can generate one for you. To make integration easier, the container does a trick with the private key by generating it with 640 permissions instead of 600 permissions. This means any member of the same group as the container can use this generated certificate-key pair. However, to make this work, you need to ensure the GID is set properly as mentioned in the specify UID/GID at runtime section above. I’ll use GID=6000 in the following example.

To generate a certificate, invoke the container with the generate-cert hostname parameter. Let’s use server.example.com:

docker run --rm -v /mycerts:/certs asifbacchus/ab-nginx generate-cert server.example.com

In this example, your self-signed certificate (fullchain.pem), private key (privkey.pem) and certification chain/bundle (chain.pem) will be saved, group-readable, in your /mycerts/ directory on the host. Remember to import the certificate on any clients or you will get warnings!

Environment variables

You can set several options simply by passing environment variables. They are pretty self-explanatory but here is a summary:

name	description	default	permitted values
TZ	Set the container time zone for proper logging.	Etc/UTC	Valid IANA TZ values
SERVER_NAMES	Space-delimited list of hostnames/FQDNs to which NGINX should respond. Must be "enclosed in quotes". This is only used by the default configuration and would not be applicable if you are using your own server blocks, unless you choose to reference the /etc/nginx/server_names.conf file. If you are using the default configuration and SSL, remember this must match your SSL certificates! The default value will only work for for HTTP connections!	"_" (this means "anything")	Valid IANA hostnames
HTTP_PORT	Port on which HTTP connections should be accepted. If you set this, make sure you set your port mapping properly! For example, if you set this to 8080 then you need to specify -p 8080:8080 or something like -p 12.34.567.89:8080:8080. In most cases, you don’t need this and should only change the host port mapping. For example -p 8080:80.	80	Valid unused ports
HTTPS_PORT	Port on which HTTPS connections should be accepted. If you set this, make sure you set your port mapping properly! For example, if you set this to 8443 then you need to specify -p 8443:8443 or something like -p 12.34.567.89:8443:8443. In most cases, you don’t need this and should only change the host port mapping. For example -p 8443:443.	443	Valid unused ports
ACCESS_LOG	Turn on/off access logging. The default format is the same as the NGINX default: combined. You can specify your own format via configuration files and use it in custom server blocks.	OFF	ON, OFF
HSTS	Activate the HSTS header. Please be sure you know what this means and that your SSL configuration is correct before enabling! The default configuration sets an HSTS max-age of 15768000s (6 months).	FALSE	Boolean: TRUE, FALSE
TLS13_ONLY	Activate the container's TLS 1.3 configuration. This is a strict TLS 1.3 implementation and does not fall back to TLS 1.2. If you still need to support TLS 1.2, then leave this turned off. The TLS 1.2 configuration does upgrade to TLS 1.3 where possible.	FALSE	Boolean: TRUE, FALSE

Shell mode

Running the container in shell mode is a great way to verify configurations, make sure everything mounted correctly or to see what the defaults are. You have two options: drop to shell before NGINX loads or after.

Drop to shell before NGINX loads

This is useful to verify where things mounted, etc. This is also useful if some configuration is causing NGINX to panic and shut down the container. Note that I’m using the --rm flag to auto-remove the container when I exit since there is no point in keeping a shell-mode instantiation around.

docker run -it --rm \
  -v ~/web:/usr/share/nginx/html \
  -v ~/nginx/config:/etc/nginx/config \
  -v ~/nginx/servers:/etc/nginx/sites \
  -v ~/certs:/certs:ro \
  asifbacchus/ab-nginx /bin/sh

Enter a running container

If you want to enter a running container and check things out:

docker exec -it ab-nginx /bin/sh

Remember this container is running Alpine Linux and the shell is ASH. You do not have all the bells and whistles of BASH! Also, many commands are run via busybox, so some things may not work exactly like you might be used to in a Debian/Ubuntu environment, for example. As a side note, ping is installed and fully functional in this container so that makes troubleshooting a little easier.

Logs

The container logs everything to stdout and stderr – in other words, the console. To see what’s going on with NGINX simply use docker’s integrated logging features from the host:

# default log lookback
docker logs ab-nginx

# last 50 lines
docker logs -n 50 ab-nginx

# show last 10 lines and follow from there in realtime (ctrl-c to stop)
docker logs -n 10 -f ab-nginx

Final thoughts

I think that's everything to get you going if you are already familiar with docker and with NGINX in general. If you need more help, please refer to the wiki. I've explained everything there in detail. Also, check out the helper scripts especially if you are deploying certificates. The scripts take care of all the docker command-lines for you so you have much less typing!

If I've forgotten anything, you find any bugs or you have suggestions, please file an issue either on my private git server or on github. Also, I am not affiliated with NGINX in any way, so please do not bother them with any issues you find with this container. Bother me instead, I actually enjoy it!

All the best and have fun!