TIG stack on G5K

Experiment Artifacts

The artifacts repository contains the E2Clab configuration files such as layers_services.yaml, network.yaml, and workflow.yaml

$ cd ~/git/
$ git clone https://gitlab.inria.fr/E2Clab/examples/monitoring-tig-g5k.git
$ cd monitoring-tig-g5k/

Defining the Experimental Environment

Layers & Services Configuration

The monitoring type is tig with the provider (machine hosting InfluxDB and Grafana) on the g5k testbed (paravance cluster). The network is shared (experiment data and monitoring data are on the same network). We use an IPv4 network. Finally, we added roles: [monitoring] in all services (e.g., MyServer, MyClientA, and MyClientB) for tig monitoring.

 1---
 2environment:
 3  job_name: monitoring-tig-g5k
 4  walltime: "00:59:00"
 5  g5k:
 6    cluster: paravance
 7    job_type: ["allow_classic_ssh"]
 8monitoring:
 9  type: tig
10  provider: g5k
11  cluster: paravance
12  network: shared
13  ipv: 4
14layers:
15- name: cloud
16  services:
17  - name: MyServer
18    environment: g5k
19    cluster: paravance
20    roles: [monitoring]
21    quantity: 1
22- name: edge
23  services:
24  - name: MyClientA
25    environment: g5k
26    cluster: paravance
27    roles: [monitoring]
28    quantity: 2
29  - name: MyClientB
30    environment: g5k
31    cluster: paravance
32    roles: [monitoring]
33    quantity: 1

Network Configuration

1networks:
2- src: cloud
3  dst: edge
4  delay: 180ms
5  rate: 1gbit
6  loss: 0.1%

Workflow Configuration

prepare installs stress on all Services.

launch runs stress on all Services.

 1# SERVER
 2- hosts: cloud.*
 3  prepare:
 4    - shell: apt install -y stress 
 5  launch:
 6    - debug:
 7        msg: "Running"
 8    - shell: stress --cpu 32 --timeout 60
 9      async: 60
10      poll: 0
11# CLIENT
12- hosts: edge.*
13  prepare:
14    - shell: apt install -y stress 
15  launch:
16    - debug:
17        msg: "Running"
18    - shell: stress --cpu 32 --timeout 60
19      async: 60
20      poll: 0

Running & Verifying Experiment Results

Find below the commands to run this example.

$ e2clab layers-services ~/git/monitoring-tig-g5k/
                         ~/git/monitoring-tig-g5k/

$ e2clab workflow ~/git/monitoring-tig-g5k/ prepare

$ e2clab workflow ~/git/monitoring-tig-g5k/ launch

You can access the Grafana service to visualize monitoring data during experiment execution. You can access it as described in the following file ~/git/monitoring-tig-g5k/20231013-150327/layers_services-validate.yaml. See more details below:

'* * * * * * * Monitoring Service (started during workflow ''launch'' step)'
'Available at: http://localhost:3000'
'Access from your local machine: ssh -NL 3000:localhost:3000 paravance-10.rennes.grid5000.fr'
'username: admin / password: admin'
../_images/figure-tig_g5k_cpu.png

Figure 1: CPU usage on G5K node ‘MyServer’.

../_images/figure-tig_g5k_mem.png

Figure 2: Memory usage on G5K node ‘MyServer’.

Wait at least one minute before finalizing the workflow and saving the monitoring data.

$ e2clab finalize ~/git/monitoring-tig-g5k/

The monitoring data will be saved at:

$ ls ~/git/monitoring-tig-g5k/20231013-150327/monitoring-data/
influxdb-data.tar.gz