This docker-compose setup creates a simulation of a cross-region Mongo replica set. The set consists of 3 instances. Two of them (m1 and m2) are installed in one "virtual datacenter" and the third (m3) is placed in a different region, with 100ms of latency.

The user observes the replica set from the m3 perspective, so they have 0ms latency connecting to m3 and 100ms to m1 and m2:

In order to setup the containers, invoke:

./init.sh

it'll use docker-compose to create 4 containers. Mongo instances are available on exposed ports 27017 (primary), 27018 (secondary) and 27019 (secondary, no latency).

One may connect to the Mongo using the ports above and the hostname of the docker-machine. If the docker runs without the docker-machine localhost will work as well.

mongo $(docker-machine ip):27017

The latency and jitter can be set using another script:

./set_latency.sh 200 10

Both values are passed in milliseconds. Invoking the script without any parameters will disable the latency.

The whole deployment can be stopped and removed with following command:

./kill.sh

• Add fakes3 support to simulate datastore latency