Skip to content

Latest commit

 

History

History
74 lines (50 loc) · 1.63 KB

setup_linux.md

File metadata and controls

74 lines (50 loc) · 1.63 KB
Raw

Setting up an agent on Linux

On Linux, it is recommended to setup the Agent using the Docker file provided in this repository. There instructions are valid for x86_64 and ARM64 (aka ARMv8 or AARCH64).

Installing Docker

  • Install docker from the automated script
curl -sSL https://get.docker.com | sh
  • Add the local account to the docker user group so that you can execute docker commands without sudo
sudo groupadd docker
sudo usermod -aG docker $USER
newgrp docker
  • Reopen the session with the account
  • Check Docker is running
docker run hello-world

Starting the Agent

  • Clone the crank repository
mkdir ~/src
cd ~/src
git clone https://github.com/dotnet/crank
  • Build and run the Agent docker image
cd ~/src/crank/docker/agent
./build.sh
./run.sh

This will build the image with all the dependencies (perfcollect, ...) and start a container named crank-agent. To stop the container, run ./stop.sh

Displaying the agent log

To display the live log, run the following command:

docker logs -f --tail 100 crank-agent

Continuous integration

In order to restart and update the agent regularly, the following cron job can be used.

  • Edit the crontab file:
crontab -e
  • Add this entry:
0 0 * * * cd [PATH_TO_CRANK]/src/crank/docker/agent; ./stop.sh; docker rm -f $(docker ps -a -q --filter "label=benchmarks"); docker system prune --all --force --volumes; git checkout -f master; git pull; ./build.sh; ./run.sh

This will stop any running agent, clean all docker images used to run benchmarks, update the GitHub repositor, build and restart the agent image.