Install Step-By-Step

This section will guide you through the FindFace Server step-by-step installation process. Follow the instructions below minding the sequence.

Tip

Standalone installation can also be done from a developer-friendly console installer or as a fully pre-configured virtual machine image.

In this section:

Prepare Packages for Installation

FindFace Enterprise Server SDK can be installed from a local repository. You can receive the FindFace Enterprise Server SDK distributable packages from your NTechLab manager. To prepare the packages for installation, do the following:

Warning

The ntech user will be automatically created at this stage. To avoid a conflict, make sure that such a user does not already exist in the system.

  1. Unpack the package with components on each designated host.

    sudo dpkg -i <findface-repo>.deb
    
  2. Add a signature key on each designated host.

    sudo apt-key add /var/findface-repo/public.key
    sudo apt-get update
    
  3. Unpack the packages with models (face, gender, age, and emotions). In the cluster environment, models are installed only on the extraction-api hosts.

    sudo dpkg -i findface-data*
    

Licensing

You receive a license file from your NTechLab manager along with the FindFace Enterprise Server SDK distributable packages. If you opt for on-premise licensing, we will also send you a Guardant USB dongle. The licensing scheme for FindFace Enterprise Server SDK is shown on the diagram below.

licensing_en

To provide the FindFace Enterprise Server SDK licensing, follow the steps below:

  1. Install and configure the local license server NTLS.
  2. If the licensable components (extraction-api, tntapi, fkvideo_detector) are installed on remote hosts, specify the NTLS host IP address in their configuration files.

To install and configure NTLS, do the following:

  1. Install the NTLS component:

    sudo apt-get update
    sudo apt-get install ntls
    

    Tip

    In the NTLS configuration file, you can change the license folder and specify your proxy server IP address if necessary. You can also change the NTLS web interface remote access settings. To open the NTLS configuration file, execute:

    sudo vi /etc/ntls.cfg
    

    If necessary, change the license folder in the license-dir parameter. By default, license files are stored at /ntech/license:

    license-dir = /ntech/license
    

    If necessary, uncomment the proxy line and specify your proxy server IP address:

    proxy = http://192.168.1.1:12345
    

    By default, you can access the NTLS web interface from any remote host (ui = 0.0.0.0:3185). To indicate that accessing the NTLS web interface must originate from a specific IP address, edit the ui parameter:

    ui = 127.0.0.1:3185
    
  2. Enable the NTLS service autostart and launch the service:

    sudo systemctl enable ntls && sudo systemctl start ntls
    
  3. Upload the license file via the NTLS web interface http://<NTLS_IP_address>:3185/#/. You can also use the NTLS web interface to consult your license information, and upgrade or extend the license.

  4. For on-premise licensing, insert the Guardant dongle into a USB port.

Important

If the licensable components (extraction-api, tntapi, fkvideo_detector) are to be installed on remote hosts, keep in mind that you have to specify the IP address of the NTLS host in their configuration files after installation.

Install Components

Now that you have prepared the FindFace Enterprise Server SDK packages and provided the licensing, install the Server components on designated host(s) according to your architecture outline.

Install facenapi

Install and configure the findface-facenapi component as follows:

  1. Install the component.

    sudo apt-get update
    sudo apt-get install python3-facenapi
    
  2. If MongoDB is installed on a remote host, specify its IP address in the findface-facenapi configuration file.

    sudo vi /etc/findface-facenapi.ini
    
    mongo_host = '192.168.113.1'
    
  3. Check if the component is runnable. To do so, invoke the findface-facenapi application by executing the command below. As the application is invoked, hold 1 minute, and if no errors display, hit Ctrl+C.

    If MongoDB is installed on the same host, execute:

    findface-facenapi
    

    If MongoDB is installed on a remote host, execute:

    sudo findface-facenapi --config=/etc/findface-facenapi.ini
    
  4. Check if the findface-facenapi service autostart at system startup is disabled.

    systemctl list-unit-files | grep findface-facenapi
    
  5. Enable the service autostart and launch the service.

    sudo systemctl enable findface-facenapi.service && sudo service findface-facenapi start
    
  6. Make sure that the service is up and running. The command will return a service description, a status (should be Active), path and running time.

    sudo service findface-facenapi status
    

Tip

You can view the findface-facenapi logs by executing:

sudo tail -f /var/log/syslog | grep facenapi

Install extraction-api

Install and configure the findface-extraction-api component as follows:

Note

The extraction-api component requires the packages with models <findface-data>.deb. Make sure they have been installed.

  1. Install the component.

    sudo apt-get update
    sudo apt-get install findface-extraction-api
    
  2. Open the findface-extraction-api.ini configuration file.

    sudo vi /etc/findface-extraction-api.ini
    
  3. If NTLS is remote, specify its IP address.

    license_ntls_server: 192.168.113.2:3133
    
  4. The model_instances parameter indicates how many extraction-api instances are used. Specify the number of instances that you purchased. The default value (0) means that this number is equal to the number of CPU cores.

    Note

    This parameter severely affects RAM consumption.

    model_instances: 2
    
  5. Enable the Extraction API service autostart and launch the service.

    sudo systemctl enable findface-extraction-api && sudo systemctl start findface-extraction-api
    
  6. Make sure that the service is up and running. The command will return a service description, a status (should be Active), path and running time.

    sudo service findface-extraction-api status
    

Tip

You can view the extraction-api logs by executing:

sudo tail -f /var/log/syslog | grep extraction-api

Install findface-upload

To store all original images which have been processed by Server, as well as such Server artifacts as face thumbnails and normalized images, install and configure the findface-upload component.

Tip

Skip this procedure if you do not want to store these data on the FindFace Enterprise Server SDK host. In this case, only face features vectors (facens) will be stored (in the MongoDB and Tarantool databases).

Do the following:

  1. Install the component:

    sudo apt-get update
    sudo apt-get install findface-upload
    
  2. By default the original images, thumbnails and normalized images are stored at /var/lib/ffupload/uploads/. You can view this folder content at http://127.0.0.1:3333/uploads/ in your browser. Make sure that this address is available (you will see Forbidden if so).

    curl -I http://127.0.0.1:3333/uploads/
    ##HTTP/1.1 403 Forbidden
    

    Important

    You will have to specify this address when configuring network.

Install tntapi

The tntapi component connects the Tarantool database and the facenapi component, transferring search results from the database to facenapi for further processing. To increase search speed, multiple tntapi shards can be created on each Tarantool host. Their running concurrently leads to a remarkable increase in performance. Each shard can handle up to approximately 10,000,000 faces. In the case of standalone deployment, you need only one shard (created by default), while in a cluster environment the number of shards has to be calculated depending on several parameters (see below).

Install tntapi standalone

Install and configure the tntapi component as follows:

  1. Install tntapi. Tarantool will be installed automatically along with it.

    sudo apt-get update
    sudo apt-get install findface-tarantool-server
    
  2. Disable the Tarantool example service autostart and stop the service.

    sudo systemctl disable tarantool@example && sudo systemctl stop tarantool@example
    
  3. For a small-scale project, the tntapi shard created by default (tarantool@FindFace) would suffice as 1 shard can handle up to 10,000,000 faces. Configuration settings of the default shard are defined in the file /etc/tarantool/instances.enabled/FindFace.lua. We strongly recommend you not to add or edit anything in this file, except the maximum memory usage (memtx_memory), the NTLS IP address required for the tntapi licensing, and the remote access setting. The maximum memory usage should be set in bytes, depending on the number of faces the shard handles, at the rate roughly 1280 byte per face.

    Open the configuration file:

    sudo vi /etc/tarantool/instances.enabled/FindFace.lua
    

    Edit the value due to the number of faces the shard handles. The value 1.2*1024*1024*1024 corresponds to 1,000,000 faces:

    memtx_memory = 1.2 * 1024 * 1024 * 1024,
    

    Specify the NTLS IP address if NTLS is remote:

    FindFace.start(“127.0.0.1”, 8001, {license_ntls_server=“192.168.113.2:3133”})
    

    With standalone deployment, you can access Tarantool by default only locally (127.0.0.1). If you want to access Tarantool from a remote host, either specify the remote host IP address in the FindFace.start section, or change 127.0.0.1 to 0.0.0.0 there to allow access to Tarantool from any IP address. In the case-study below, you allow access only from 192.168.113.10:

    FindFace.start("192.168.113.10", 8001, {license_ntls_server=“192.168.113.2:3133”})
    

    Now you allow access from any IP address:

    FindFace.start("0.0.0.0", 8001, {license_ntls_server=“192.168.113.2:3133”})
    
  4. Configure the tntapi shard to autostart and start the shard.

    sudo systemctl enable tarantool@FindFace && sudo systemctl start tarantool@FindFace
    
  5. Retrieve the shard status. The command will return a service description, a status (should be Active), path and running time.

    sudo systemctl status tarantool@FindFace
    

    Tip

    You can view the tntapi logs by executing:

    sudo tail -f /var/log/tarantool/FindFace.log
    
  6. The tntapi.json file containing the tntapi shard parameters is automatically installed along with tntapi into the /etc/ folder.

    Important

    You will have to uncomment the path to tntapi.json when configuring network.

Install tntapi cluster

Install and configure the tntapi component as follows:

  1. Install tntapi on designated hosts. Tarantool will be installed automatically along it.

    sudo apt-get update
    sudo apt-get install findface-tarantool-server
    
  2. Create tntapi shards on each tntapi host. To learn how to shard, let’s consider a case-study of a cluster environment containing 4 database hosts. We’ll create 4 shards on each.

    Important

    When creating shards in large installations, observe the following rules:

    1. 1 tntapi shard can handle approximately 10,000,000 faces.
    2. The number of shards on a single host should not exceed the number of its physical processor cores minus 1.
  3. Disable the Tarantool example service autostart and stop the service. Do so for all the 4 hosts.

    sudo systemctl disable tarantool@example && sudo systemctl stop tarantool@example
    
  4. Disable the shard created by default. Do so for all the 4 hosts.

    sudo systemctl disable tarantool@FindFace
    
  5. Write a bash script shard.sh that will automatically create configuration files for all shards on a particular host. Do so for the 4 hosts. Use the following script as a base for your own code. The exemplary script creates 4 shards listening to the ports: tntapi 33001..33004 and http 8001..8004.

    Important

    The script below creates configuration files based on the default configuration settings stored in the file /etc/tarantool/instances.enabled/FindFace.lua. We strongly recommend you not to add or edit anything in the default file, except the maximum memory usage (memtx_memory) and the NTLS IP address required for the tntapi licensing. The maximum memory usage should be set in bytes for each shard, depending on the number of faces a shard handles, at the rate roughly 1280 byte per face.

    Open the configuration file:

    sudo vi /etc/tarantool/instances.enabled/FindFace.lua
    

    Edit the value due the number of faces a shard handles. The value 1.2*1024*1024*1024 corresponds to 1,000,000 faces:

    memtx_memory = 1.2*1024*1024*1024,
    

    Specify the NTLS IP address if NTLS is remote:

    FindFace.start(“127.0.0.1”, 8001, {license_ntls_server=“192.168.113.2:3133”})
    
    #!/bin/sh
    set -e
    
    for I in `seq 1 4`; do
           TNT_PORT=$((33000+$I)) &&
           HTTP_PORT=$((8000+$I)) &&
           sed "
                   s#/opt/ntech/var/lib/tarantool/default#/opt/ntech/var/lib/tarantool/shard_$I#g;
                   s/listen = .*$/listen = '127.0.0.1:$TNT_PORT',/;
                   s/\"127.0.0.1\", 8001,/\"0.0.0.0\", $HTTP_PORT,/;
           " /etc/tarantool/instances.enabled/FindFace.lua > /etc/tarantool/instances.enabled/FindFace_shard_$I.lua;
    
           mkdir -p /opt/ntech/var/lib/tarantool/shard_$I/snapshots
           mkdir -p /opt/ntech/var/lib/tarantool/shard_$I/xlogs
           mkdir -p /opt/ntech/var/lib/tarantool/shard_$I/index
           chown -R tarantool:tarantool /opt/ntech/var/lib/tarantool/shard_$I
           echo "Shard #$I inited"
    done;
    

    Tip

    Download the exemplary script.

  6. Run the script from the home directory.

    sudo sh ~/shard.sh
    
  7. Check the configuration files created.

    ls /etc/tarantool/instances.enabled/
    
    ##example.lua FindFace.lua FindFace_shard_1.lua FindFace_shard_2.lua FindFace_shard_3.lua FindFace_shard_4.lua
    
  8. Launch all the 4 shards. Do so on each host.

    for I in `seq 1 4`; do sudo systemctl enable [email protected]_shard_$I; done;
    for I in `seq 1 4`; do sudo systemctl start [email protected]_shard_$I; done;
    
  9. Retrieve the shards status.

    sudo systemctl status tarantool@FindFace*
    

    You should get the following output:

    tarantool@FindFace_shard_3.service - Tarantool Database Server
    Loaded: loaded (/lib/systemd/system/tarantool@.service; disabled; vendor preset: enabled)
    Active: active (running) since Tue 2017-01-10 16:22:07 MSK; 32s ago
    ...
    tarantool@FindFace_shard_2.service - Tarantool Database Server
    Loaded: loaded (/lib/systemd/system/tarantool@.service; disabled; vendor preset: enabled)
    Active: active (running) since Tue 2017-01-10 16:22:07 MSK; 32s ago
    ...
    tarantool@FindFace_shard_1.service - Tarantool Database Server
    Loaded: loaded (/lib/systemd/system/tarantool@.service; disabled; vendor preset: enabled)
    Active: active (running) since Tue 2017-01-10 16:22:07 MSK; 32s ago
    ...
    tarantool@FindFace_shard_4.service - Tarantool Database Server
    Loaded: loaded (/lib/systemd/system/tarantool@.service; disabled; vendor preset: enabled)
    Active: active (running) since Tue 2017-01-10 16:22:07 MSK; 32s ago
    ...
    

    Tip

    You can view the tntapi logs by executing:

    sudo tail -f /var/log/tarantool/FindFace_shard_{1,2,3,4}.log
    
  10. On the findface-facenapi host, create a file tntapi_cluster.json containing the addresses and ports of all the shards. Distribute available shards evenly over ~1024 cells in one line. Click here to see the file for 4 hosts with 4 shards on each.

    Tip

    You can create tntapi_cluster.json as follows:

    1. Create a file that lists all the shards, each shard with a new line (click here to view the example).

      sudo vi s.txt
      
    2. Run the script below (click here to view the script). As a result, a new file tntapi_cluster.json will be created and contain a list of all shards distributed evenly over 1024 cells.

    cat s.txt | perl -lane 'push(@s,$_); END{$m=1024; $t=scalar @s;for($i=0;$i<$m;$i++){$k=int($i*$t/$m); push(@r,"\"".$s[$k]."\"")} print "[[".join(", ",@r)."]]"; }' > tntapi_cluster.json
    
  11. Move tntapi_cluster.json to the directory /etc/.

    Important

    You will have to uncomment and specify the path to tntapi_cluster.json when configuring network.

Configure Network

After you install the FindFace Server components, configure their interaction with each other. Do the following:

  1. Open the findface-facenapi.ini configuration file:

    sudo vi /etc/findface-facenapi.ini
    
  2. Uncomment and/or edit the settings to align with your network specifications, substituting the suggested values with actual location:

    ffupload_url = 'http://127.0.0.1:3333'
    mongo_host = '127.0.0.1'
    extraction_api_url = 'http://127.0.0.1:18088'
    tntapi_servers_file = '/etc/tntapi.json'
    

    Warning

    The findface-facenapi.ini content must be correct Python code.

    Note

    Do not specify ffupload_url if the findface-upload component is not installed.

  3. By default, if one or several tntapi shards are out of service during face identification, findface-facenapi returns an error. If necessary, uncomment the tntapi_ignore_search_error parameter and assign it True. In this case findface-facenapi will use available tntapi shards to obtain face identification results, indicating the number of available servers vs the total number of servers in the response:

    tntapi_ignore_search_errors = True
    
  4. Restart all the FindFace Enterprise Server SDK services and nginx (for findface-upload) on the relevant host(s).

    sudo service 'findface*' restart
    sudo service nginx restart
    
  5. Check the services status. The command will return the services description, status (should be Active), path and running time.

    sudo service 'findface*' status
    sudo service nginx status