The next major update to this container will default to running UniFi as a non root user.
If you would like to continue to running as a root user then simply add the environment variable RUNAS_UID0
with the value true to your container configuration. It is preffered you set ownership of your unifi data and logs
to the proper uid/gid (default is 999/999) and run as a restricted user.
The beta readme has more details on the changes.
| Tag | Description |
|---|---|
latest, stable, unifi-5.5 |
Tracks UniFi stable version - 5.5.20 as of 2017-07-31 |
oldstable, unifi-5.4 |
Tracks UniFi Old Stable version - 5.4.19 as of 2017-07-31 |
sc |
Tracks UniFi "Stable Candidate", this tag however may not always cleanly upgrade and it is better to track the direct sc tags ie : 5.6.18-sc instead. |
These tags generally track the UniFi APT repository. We do lead the repository a little when it comes to pushing the latest version. The latest version gets pushed when it moves from stable candidate to stable instead of waiting for it to hit the repository.
In adition to these tags you may tag specific versions as well, for example jacobalberty/unifi:5.4.19 will get you unifi 5.4.19 no matter what the current version is.
Stable candidates now exist both under the sc tag and for tags with the extension -sc ie jacobalberty/unifi:5.6.18-sc. It is advised to use the specific versions as the sc tag may jump from 5.6.x to 5.5.x then back to 5.6.x as new stable candidates come out.
This is a containerized version of Ubiqiti Network's Unifi Controller version 5.
The following options may be of use:
- Set the timezone with
TZ - Bind mount the
dataandlogvolumes
It is suggested that you include --init to handle process reaping Example to test with
mkdir -p unifi/data
mkdir -p unifi/logs
docker run --rm --init -p 8080:8080 -p 8443:8443 -p 3478:3478/udp -p 10001:10001/udp -e TZ='Africa/Johannesburg' -v ~/unifi/data:/var/lib/unifi -v ~/unifi/logs:/var/log/unifi --name unifi jacobalberty/unifi:unifi5The default example requires some l3 adoption method. You have a couple options to adopt.
The quickest one off method is to ssh into the access point and run the following commands
mca-cli
set-inform http://<host_ip>:8080/inform
You can see more options on the (UniFi website)[https://help.ubnt.com/hc/en-us/articles/204909754-UniFi-Layer-3-methods-for-UAP-adoption-and-management]
You can also enable layer 2 adoption through one of two methods.
If you launch the container using host networking (With the --net=host parameter on docker run) Layer 2 adoption works as if the controller is installed on the host.
It is possible to configure the macvlan driver to bridge your container to the host's networking adapter. Specific instructions for this container are not yet available but you can read a write-up for docker at http://collabnix.com/docker-17-06-swarm-mode-now-with-macvlan-support/
There is now a new beta branch on github to support easier building of betas. This branch does not exist on the docker hub at all, and must be built from the git repository.
You simply build and pass the build argument PKGURL with the url to the .deb file for the appropriate beta you wish to build. I believe
this will keep closest with the letter and spirit of the beta agreement on the unifi forums while still allowing relatively easy access to the betas.
This build method is the method I will be using for my own personal home network to test the betas on so it should remain relatively well tested.
If you would like to submit a new feature for the images the beta branch is probably a good one to apply it against as well.
I will be cleaing up the Dockerfile under beta and gradually pushing out the improvements to the other branches. So any major changes
should apply cleanly against the beta branch.
The command line is pretty simple:
docker build -t unifi-beta --build-arg PKGURL=https://dl.ubnt.com/unifi/5.5.24/unifi_sysvinit_all.deb "https://github.com/jacobalberty/unifi-docker.git#beta"
Simply replace the url to the debian package with the version you prefer.
This is just as easy when using version 2 of the docker-compose.yml file format.
Under your containers service definition instead of using image: jacobalberty/unifi use the following:
build:
context: https://github.com/jacobalberty/unifi-docker.git#beta
args:
PKGURL: https://dl.ubnt.com/unifi/5.5.24/unifi_sysvinit_all.deb
Once again, simply change PKGURL to point to the package you would like to use.
Configuration data
Log files
Run information
TimeZone. (i.e America/Chicago)
used to set max thread stack size for the JVM
81A0Ex:
--env JVM_MAX_THREAD_STACK_SIZE=1280k
While micro-service patterns try to avoid running multiple processes in a container, the unifi5 container tries to follow the same process execution model intended by the original debian package and it's init script, while trying to avoid needing to run a full init system.
dumb-init has now been removed. Instead it is now suggested you include --init in your docker run command line.
If you are using docker-compose you can accomplish the same by making sure you use version 2.2 of the yml format and add init: true to your service definition.
unifi.sh executes and waits on the jsvc process which orchestrates running the controller as a service. The wrapper script also traps SIGTERM to issue the appropriate stop command to the unifi java com.ubnt.ace.Launcher process in the hopes that it helps keep the shutdown graceful.
You may now place init scripts to be launched during the unifi startup in /usr/local/unifi/init.d to perform any
actions unique to your unifi setup. An example bash script to set up certificates is in /usr/unifi/init.d/import.sh.
To use custom SSL certs, you must map a volume with the certs to /var/cert/unifi
They should be named:
cert.pem # The Certificate
privkey.pem # Private key for the cert
chain.pem # full cert chain
For letsencrypt certs, we'll autodetect that and add the needed Identrust X3 CA Cert automatically.
Future work?
- Don't run as root (but Unifi's Debian package does by the way...)
- Possibly use Debian image with systemd init included (but thus far, I don't know of an official Debian systemd image to base off)