Skip to content

Latest commit

 

History

History
123 lines (96 loc) · 4.52 KB

DEOPS-load_balancing.md

File metadata and controls

123 lines (96 loc) · 4.52 KB
Raw

Task - Ingress Load Balancing

In this task, you will observe the behaviour of the built in load balancing abilities of the Ingress network

Pre-requisites

  1. UCP installed with 2 worker nodes

Step 1 - Deploy ElasticSearch service

  1. Deploy a Service using the elasticsearch:2.3 image and specify the following details:
    • **Service Name: ** elastic-load-balanced
    • Published Ports - Publish the service port 9200 onto the public port of 9200.
    • Mode: Replicated
    • Scale: 3
  2. Once the Service is deployed check the Task tab to confirm that there are 3 containers running

Step 2 - Observe load-balancing and Scale

  1. SSH into the ucp-controller node

  2. Run curl localhost:9200 and observe the output. You should see something similar to the following:

    ubuntu@ucp-controller:~$ curl localhost:9200
{
  "name" : "Entropic Man",
  "cluster_name" : "elasticsearch",
  "version" : {
    "number" : "2.3.5",
    "build_hash" : "90f439ff60a3c0f497f91663701e64ccd01edbb4",
    "build_timestamp" : "2016-07-27T10:36:52Z",
    "build_snapshot" : false,
    "lucene_version" : "5.5.0"
  },
  "tagline" : "You Know, for Search"
}

  3. Take note of the value of the name attribute. In this example, our name is Entropic Man. ElasticSearch instances use a unique name to identify themselves. So each one of our ElasticSearch container should have a different name.

  4. Run curl localhost:9200 again. What can you observe? In our example we have the following:

    ubuntu@ucp-controller:~$ curl localhost:9200
{
  "name" : "Horus",
  "cluster_name" : "elasticsearch",
  "version" : {
    "number" : "2.3.5",
    "build_hash" : "90f439ff60a3c0f497f91663701e64ccd01edbb4",
    "build_timestamp" : "2016-07-27T10:36:52Z",
    "build_snapshot" : false,
    "lucene_version" : "5.5.0"
  },
  "tagline" : "You Know, for Search"
}

  5. Notice how everything was the same, except the name. This shows us that the UCP routing mesh has sent our 2nd request over to a different container, since the name was different.

  6. Repeat the command two more times. What can you observe?

    ubuntu@ucp-controller:~$ curl localhost:9200
{
  "name" : "Horus",
  "cluster_name" : "elasticsearch",
  "version" : {
    "number" : "2.3.5",
    "build_hash" : "90f439ff60a3c0f497f91663701e64ccd01edbb4",
    "build_timestamp" : "2016-07-27T10:36:52Z",
    "build_snapshot" : false,
    "lucene_version" : "5.5.0"
  },
  "tagline" : "You Know, for Search"
}
ubuntu@ucp-controller:~$ curl localhost:9200
{
  "name" : "Entropic Man",
  "cluster_name" : "elasticsearch",
  "version" : {
    "number" : "2.3.5",
    "build_hash" : "90f439ff60a3c0f497f91663701e64ccd01edbb4",
    "build_timestamp" : "2016-07-27T10:36:52Z",
    "build_snapshot" : false,
    "lucene_version" : "5.5.0"
  },
  "tagline" : "You Know, for Search"
}

    You should see one new name and then on the 4th request it should revert back to the name of the first container. In this example that name is Entropic Man

  7. Now let's scale the number of Tasks for our elastic-load-balanced service. We will add 3 more Tasks to the service so that we have a total of six ElasticSearch containers.

    a) First, click on Resources and go to the Services page.

    b) Then click on the elastic-load-balanced service that we've created.

    c) Click on the Scheduling Tab in the service details and edit the Scale field to 6 as shown below:

    d) Once you have edited the field, click on the Tick symbal and then click Save Changes on the top right.

  8. Now click on the Tasks Tab of the elastic service and verify that there are 6 tasks running

  9. Now switch back to you SSH terminal and run curl localhost:9200 multiple times again. You should be able to observe some new names.

Step 3 - Ingress Routing Mesh

Remember the Nginx_service we deployed in the previous Task? Try and hit the NGINX server on any node in your cluster.

  1. Open a browser tab and specify the public domain or IP address, followed by port 8080 of any of your UCP nodes. Try at least the following nodes:

    • ucp-controller
    • ucp-node-0
    • ucp-node-1
    • ucp-managers-0
    • ucp-managers-1

  2. You should be able to observe the NGINX welcome page on all of them, even though we are only running one NGINX container.