Minor improvements on the Dockerfile

[ema-pe]

These are some notes about the Dockerfile to build the dfaas node that I collected during some manual deployments:

1. Use the hack/install.sh script provided from faasd, updated with the given version, and not the local one under docker/files/faasd/hack/install.sh:

   dfaas/docker/dfaas-node.dockerfile

   Lines 29 to 31 in 034794c

   	COPY files/faasd/hack/install.sh ./hack/install.sh
   	RUN chmod +x ./hack/install.sh
   	RUN ./hack/install.sh

2. The faasd and faasd-provider units are already enabled by default when installing faasd using the hack/install.sh script:

   dfaas/docker/dfaas-node.dockerfile

   Lines 35 to 36 in 034794c

   	RUN systemctl enable faasd.service
   	RUN systemctl enable faasd-provider.service

3. When upgrading cAdvisor to v0.51.0, use the direct link (see release assets) and move the binary to a better location (like /opt/cadvisor):

   dfaas/docker/dfaas-node.dockerfile

   Lines 38 to 42 in 034794c

   	# Add cAdvisor to monitor containers
   	RUN wget https://github.com/google/cadvisor/releases/download/v0.44.0/cadvisor
   	RUN chmod +x cadvisor
   	COPY files/faasd/cadvisor.service /etc/systemd/system/cadvisor.service
   	RUN systemctl enable cadvisor.service

4. Same issue as the previous point with Prometheus node exporter:

   dfaas/docker/dfaas-node.dockerfile

   Lines 44 to 48 in 034794c

   	# Add Prometheus node exporter to monitor node
   	RUN wget https://github.com/prometheus/node_exporter/releases/download/v1.3.1/node_exporter-1.3.1.linux-amd64.tar.gz
   	RUN tar xvfz node_exporter-1.3.1.linux-amd64.tar.gz && rm node_exporter-1.3.1.linux-amd64.tar.gz
   	COPY files/faasd/node-exporter.service /etc/systemd/system/node-exporter.service
   	RUN systemctl enable node-exporter.service

5. For the forecaster component, install it in /opt/dfaas-forecaster instead of the root directory (this requires updating the config_constants.py file). In the repository, it should be moved outside the docker directory, as it is a component not directly related to docker. Do not install the dependencies via pip when running as root, use apt instead (or create a virtual environment, see this for how to update the unit file), as it may conflict with the system-wide packages.

   dfaas/docker/dfaas-node.dockerfile

   Lines 61 to 67 in 034794c

   	### DFaaS Forecaster
   	WORKDIR /forecaster
   	COPY forecaster/ .
   	RUN pip install "fastapi[all]" scikit-learn lightgbm joblib pandas numpy
   	COPY files/forecaster/forecaster.service /etc/systemd/system/forecaster.service
   	RUN systemctl enable forecaster.service
   	### End DFaaS Forecaster

6. Same for the dfaasagent component: it is better to install it in /opt/dfaas-agent instead of /agent, because it is a known directory in FHS. This requires some implementation changes in the agent.

   dfaas/docker/dfaas-node.dockerfile

   Lines 69 to 77 in 034794c

   	### Agent
   	WORKDIR /agent
   	COPY files/dfaasagent/dfaasagent.service /etc/systemd/system/dfaasagent.service
   	RUN systemctl enable dfaasagent.service
   	COPY --from=dfaas-agent-builder:latest /go/src/dfaasagent/agent/loadbalancer/haproxycfgrecalc.tmpl ./haproxycfgrecalc.tmpl
   	COPY --from=dfaas-agent-builder:latest /go/src/dfaasagent/agent/loadbalancer/haproxycfgnms.tmpl ./haproxycfgnms.tmpl
   	COPY --from=dfaas-agent-builder:latest /go/src/dfaasagent/agent/loadbalancer/group_list.json ./group_list.json
   	COPY --from=dfaas-agent-builder:latest /go/src/dfaasagent/dfaasagent ./dfaasagent
   	### End Agent