TTM-backend

A backend to serve travel time matrices as geodata. Matrices are served as catchment polygons in geojson format, precompressed with gzip. There is one file for every grid cell / travel mode combination.

This repository holds only what is necessary to serve the ready-made data. See preprocessing for producing the data.

Overview

Technology

• Nginx as a server
• Kubernetes (OpenShift) for deploying to csc's rahti container cloud

Server

The approach is to serve geojsons directly from the filesystem as static files. A modified nginx image is used for this (See Dockerfile). It uses a custom config file and makes some additional adjustments to run nginx without root. This is needed to make the image run on OpenShift.

Deployment

The kubernetes configuration is in manifest.yml. It defines:

• A Deployment that creates and runs a ReplicaSet of server pods (each running the custom nginx image)
• A PersistentVolumeClaim for persistent data storage (mounted to all replicas)
• A Service for exposing the deployment network
• A Route for routing traffic from the internet to the Service

Developing / deploying the backend

Locally

A container engine such as podman or docker is required.

Run

docker compose up

to start the service.

The server serves files from the geojson/ directory on http://localhost:8080/. You can copy the contents of geojson-example/ to geojson/ in order to get example data for a single grid cell.

Deploying to rahti

The command line tool oc is required for interacting with rahti.

After authenticating you can create a deployment with:

oc create -f manifest.yml

To get data to rahti, run:

oc rsync </local/dir> <POD-name>:/usr/share/nginx/geojson

You can get the pod name with oc get pods. Confusingly, it does not matter which pod you clone to, since all replicas share the persistent volume.

You can delete everything except the PersistentVolume with:

oc delete all --selector app=ttm

The volume stays intact, and all data is ready for next deploy.