Initial Server Setup

AWS Servers

This code base can be used on any server (in-house or remote/virtual). Four AWS instances are currently in use:

• aws-model01
• aws-model03
• aws-model04
• aws-model05
• aws-model06
• aws-model07
• aws-model08
• aws-model09

These machines images were created from aws-model01, and include identical verisons of software. The original installers used for these are available at R:\Urbansim\integration. The base sofware versions are as follows:

• Anaconda2-4.3.1-Windows-x86_64.exe
• Git-2.18.0-64-bit.exe
• Emme-4.3.5-x64.exe

Other default settings include preset environment variables, as follows:

• OPUS_HOME=E:\opusgit\urbansim_data
• OPUS_DATA_PATH=E:\opusgit\urbansim_data\data
• PYTHONPATH=E:\opusgit\urbansim

When building machines from the aws-model01 image, these variables should be defined automatically. Note that a user may need to log in once before a newly-imaged machine is available on the local network.

Amazon cloud servers are managed remotely. See Piset for account access. From this portal, different instance types (memory capacity, processor capability) can be selected. Machines may also be started/stopped/paused. When not in immediate use, all servers should be stopped or in use as a cost-effective micro version. The current instance type recommendations for modeling are as follows:

• full integrated runs: r5.8xlarge
• Soundcast only: r5.8xlarge
• Urbansim only: c5.4xlarge

To make these changes, visit the EC2 tab and view instances. Right click on the machine to be used and select Instance State or Change Instance Type. AWS server are accessed the same as for local servers, using the shared modeler account.

Virtual Environment

A uniform virtual environment called model is used for Urbansim, Soundcast, and integrated modeling. The latest version is stored within this repository at spec\model-spec.yml. This is also available on the network drive at R:\Urbansim\integration\model-spec.yml.

The virtual environment should be automatically installed on new machine images, but may need to be manually installed on certain servers, or may need to be reinstalled to ensure latest version is used. To create the virtual environment, open the Anaconda prompt from the server and enter:

conda env create -f R:\urbansim\integration\model-spec.yml

Activate this environment before beginning runs from the Anaconda prompt with activate model

Set Drive Permissions

For a new machine, local drives must be given full permissions for all users.

• right-click on the E drive and select the Security tab
• Edit permissions to allow 'Everyone' and 'Users' to Modify files. Under Advanced Security Settings in the main Security tab, ensure that the controls apply to "This folder, subfolder, and files."
• Additionally, ensure that the Anaconda prompt is *not run as administrator. To set the prompt to not run as admin by default, right-click the icon to get to the properties -> Shortcut -> Advanced -> Uncheck "Run as administrator".

Emme

Emme is set up for all current servers 1,3,4, and 5. The following instructions are for setting up new machines. New images should include the software installed, but access and authorization must be granted before the software can be used for modeling.

Authorization Files and Keys

• Open INRO Software Manager (running in background - a hat icon in the right-bottom of the screen, or open from Emme install directory)

  • On Authorizations tab, Choose Emme -> Add Authorization -> "I'm not online"

  • Enter activation code from Kris

  • Save the authorization request to a network drive and open on your local machine (this is required because the authorization request will not work on AWS machines through the web, and must be generated from the machine that requires authorization.)

  • Open the request file from a local machine and enter Kris's credentials. Kris will receive an email of the authorization file, which should be copied to the server.

  • Reopen the INRO Software Manager -> Emme -> Add Authorization -> "I already have an authorization file" and select the copied authorization file.

  • If a license is not available see Step 3 from this reference. Licenses can be used by a single machine at a time and must first be available (deactivated) from one of the on-site servers before they are accessible for AWS machines. The following list details the key mapping between on-site servers and AWS machines:

    • aws-model01 | modelsrv2
    • aws-model03 | modelsrv1
    • aws-model04 | modelsrv6
    • aws-model05 | modelsrv8
    • aws-model06 | modelsrv5
    • aws-model07 | modelsrv7
    • aws-model08 | modelsrv4
    • aws-model09 | modelsrv8

Set Emme Python Settings

Open Emme Desktop to ensure the correct virtual environment is accessible to Emme

• Tools -> Application Options -> Modeller
• Set Python path to: C:\Anaconda\envs\model, select Apply and 'Install Modeller Package'

• Set Toolboxes to include:
  • %<$ProjectPath>%\standard.mtbx
  • %<EmmePath>%\toolboxes\manual.mtbx

Troubleshooting

To re-start an Urbansim run (integrated or non-integrated) at a specific year:

• Navigate to the opusgit\urbansim\opus_core\tools directory to find the restart_run.py script
• Restart a run from a terminal. For example, for run run_32.run_2016_10_17_15_00
  • python restart_run 32 2031 -p psrc_parcel