Accelerator Options

Hi Brian,

We have recently published a document about the Accelerator Manual Installation. You can find more about that here: https://forum.unity.com/threads/accelerator-documentations.789593/

When you launch the accelerator manually, it’s preferable to use the downloaded binary unity-accelerator instead of unity-accelerator-launch, which is our script for the installer wizard to interact with the Accelerator binary file.

Regarding your questions,

1. The HTTPPort is mainly used if you enable Accelerator for Unity Collaborate. It is also used for serving /metrics endpoint, which provides Prometheus metric reports. The ProtobufPort is used if you enable Accelerator for Asset Import Pipeline Version 2. You can specify the port to use by providing the flag -http-port or -protobuf-port when you run the binary.

# Accelerator will listen on port 1234 for Protobuf requests and port 5678 for HTTP requests

./unity-accelerator run -protobuf-port 1234 -http-port 5678

2. CacheDir is a different storage directory where we store the cache files, distinct from where the configuration and log files, .etc are stored. This is different from the persisted storage location provided through a flag described below.

Under normal circumstances, when the Accelerator was first running, configuration and log files are generated but not the cache folder (which will be called cache), which is only created when the Editor sends WRITE requests to the Accelerator.​

From what you described, I assumed unity-accelerator-cache is the storage location you provided during setup of installing the Accelerator. One way to change the cache location is to change the value of CacheDir as mentioned above. Another way is to use the binary file to register and run the Accelerator in a different location as below. If you don’t choose where persisted data will be stored, the default would be the current directory.

# Accelerator will register to use Asset Import Pipeline Version 2 and create a configuration at PersonalCache directory

./unity-accelerator register -persist PersonalCache adbv2



# Accelerator will run and use PersonalCache directory as storage location. The storage location needs to have a configuration file for the accelerator to run.

./unity-accelerator run -persist PersonalCache

3. HTTPPort and ProtobufPort fields are the configured ports that you can set through the flags mentioned above. LastUsedHTTPPort and LastUsedProtobufPort are the actual ports that the Accelerator were listening on. When the configured ports are not set or have zero values, the Accelerator would choose a random available ports to listen on, which explains why the last used ports were assigned random values when you start the Accelerator.​

 

Could you tell me what OS you are running the Accelerator on?

1. On Linux, we use service console utility to control the unity-accelerator service. From your error, it seems like the service application could not be found on your system. Please refer to these manuals for more information about service and systemctl. And to successfully complete the GUI setup, besides ensuring service utility is on the system, please run the .run program as root. This is because the installer wizard needs root access to create new service through the above service application.
2. The accelerator runs as two processes, a parent process and a child process. The parent process simply keeps the child process running -- it will relaunch the child if it exits. This is to facilitate automatic updates because, on Windows, a running executable path cannot be replaced.
   
   The parent process will always run with the installed executable. What the child process runs depends. At first, it will be running the installed executable as well. When a new version is released, it will download it to unity-accelerator-next in the --persist directory, but it won't run it right away.
   
   When the accelerator is restarted, or it reaches its once-a-day maintenance window, it will rename any unity-accelerator-next executable to unity-accelerator, and the child will then run that unity-accelerator executable in the --persist directory (if it exists, otherwise it falls back to the installed executable).
   
   You can disable auto-updates with the --no-auto-updates command-line option (see 'unity-accelerator run --help' for the available options) and remove any unity-accelerator or unity-accelerator-next executable in the --persist directory. In this mode, there will still be parent and child processes, but the downloading of new versions won't happen anymore.
   

 

@gp-lhota , sorry about the confusion regarding CacheDir. The CacheDir is where the folder “cache” will be created and on initial registration, this would be the same directory where the configuration and log files, .etc are stored. If you modify the configuration to change the value of CacheDir, the folder “cache” will be created in the new location. The folder “cache” is not created until the Editor writes data to the Accelerator. So, one way to test whether the Editor is actually working with the Accelerator is to check if the folder “cache” is created and if there is anything under that “cache” directory.

For your connection with the Accelerator in the Hyper-V, I suspect there may be a connection issue between your local machine with the Hyper-V. There are some ways to test the connection to the Accelerator:

• To test if you can connect to the Accelerator, you can send a request from your local machine to [Accelerator_IP]:[Accelerator_LastUsedHTTPPort]/api/health or [Accelerator_IP]:[Accelerator_LastUsedHTTPPort]/metrics. Here, Accelerator_IP could be the Hyper-V’s IP address and the ports can be found in the configuration.
  
  
• To test if files can be written to the Accelerator, you can use Accelerator built-in tool to send a WRITE request to the running Accelerator:

1. Running Accelerator in the Hyper-V
2. Making sure accelerator is running by sending requests to the above endpoint from your local machine (not inside Hyper-V)
3. To use the following tool, you would need to download Accelerator Executable for Windows to your local machine by following the PDF in this thread: https://forum.unity.com/threads/accelerator-documentation.789593/
4. Use tool put command on local machine to send a file to the Accelerator.
   
   

   ./unity-accelerator tool put [Accelerator_IP]:[Accelerator_LastUsedProtobufPort] namespace key [ANY_TEXT_FILE]

   
   
5. You can then use tool get command on local machine to fetch the content of the file and make sure there’s no difference.
   
   

   ./unity-accelerator tool get [Accelerator_IP]:[Accelerator_LastUsedProtobufPort] namespace key

   
   
6. If you run Accelerator on Hyper-V with -debug flag, you should see debug log with messages “handled putRequest” and “handled getRequest” in the log file when you use the above tools to test sending files to Accelerator.

If the above did not work for you and the Accelerator is running fine inside the Hyper-V (health and metrics endpoint are reachable), there may be some setup you need to do to allow your local machine to connect to applications running inside the Hyper-V with corresponding ports.

Also, we have noticed that the Check connection button was not working and this should be fixed when 2019.3 is released.

 

Hey @io-games, unfortunately we do not support Unix.

Unity Accelerator currently supports:

• Linux (Ubuntu 16.04, Ubuntu 18.04, or CentOS 7)
  For Linux you must install Acccelerator as a root user.
• Windows Server 2008R2 / Windows 7 or higher (64bit)
• Mac OS X 10.12 or higher (64bit)

Thank for this feedback - we'll add it to future OSes to support.