How can we see which nodes are primary or replica, in aws elasticache redis-cluster?

Question

At the below link:

http://docs.aws.amazon.com/AmazonElastiCache/latest/UserGuide/ClientConfig.ReplicationGroup.html

following content is present:

REPLICATIONGROUP  my-repgroup  My replication group  available

  CLUSTERID  my-redis-primary
  CLUSTERID  my-replica-1
  NODEGROUP  0001  my-repgroup.f310xz.ng.0001.cache.amazonaws.com  6379  available
        NODEGROUPMEMBER  my-redis-primary    0001  my-redis-primary.f310xz.0001.cache.amazonaws.com   6379  us-west-2a  primary
        NODEGROUPMEMBER  my-replica-1        0001  my-replica-1.f310xz.0001.cache.amazonaws.com       6379  us-west-2b  replica
Connecting to Clusters in a Replication Group (ElastiCache API)

In the above, to the extreme right
There is 'primary' and 'replica'

as of today, in the aws console,
when i give the 'describe-replication-groups' command
i don't see the 'primary' and 'replica'

instead i see something like below:

"NodeGroupMembers": [
    {
        "PreferredAvailabilityZone": "us-west-2b", 
        "CacheNodeId": "0001", 
        "CacheClusterId": "ec-redis-cluster1-0001-001"
    }, 
    {
        "PreferredAvailabilityZone": "us-west-2b", 
        "CacheNodeId": "0001", 
        "CacheClusterId": "ec-redis-cluster1-0001-002"
    }, 
    {
        "PreferredAvailabilityZone": "us-west-2b", 
        "CacheNodeId": "0001", 
        "CacheClusterId": "ec-redis-cluster1-0001-003"
    }
]

i tried various commands, could not find.

in what way, can i see the details of if some shard/node is primary/replica

thanks in advance

Answer 1

echo "cluster nodes" | redis-cli -c -h redis-cluster | grep master

Answer 2

AWS Elasticache Redis setup when created as a multi node setup has 2 modes - cluster mode enabled and disabled as described here. This is important to the question at hand.

If your Elasticache cluster is in cluster mode disabled then doing

aws elasticache describe-elasticache-replication-groups --replication-group-id test-1 

will show output similar to this:

{
  "CurrentRole": "replica",
    "PreferredAvailabilityZone": "us-west-1c",
    "CacheNodeId": "0001",
    "ReadEndpoint": {
      "Port": 6379,
      "Address": "xxxx"
    },
    "CacheClusterId": "xxxx"
},
{
  "CurrentRole": "primary",
  "PreferredAvailabilityZone": "us-west-1c",
  "CacheNodeId": "0001",
  "ReadEndpoint": {
    "Port": 6379,
    "Address": "xxxx"
  },
  "CacheClusterId": "xxxx"
}
                      

which helps you determine the target node's primary or replica role. The key CurrentRole is not available in AWS Elasticache cluster mode enabled where doing

aws elasticache describe-elasticache-replication-groups --replication-group-id test-2 

will show output similar to:

{
    "PreferredAvailabilityZone": "us-west-1c",
        "CacheNodeId": "0001",
        "CacheClusterId": "xxxxx"
},
{
    "PreferredAvailabilityZone": "us-west-1c",
    "CacheNodeId": "0001",
    "CacheClusterId": "xxxxxxx"
},

The above output doesn't have CurrentRole or any other corresponding key that gives insight into the node's role. In this case you have 2 options:

1. If you have access to port 6379 of the AWS Elasticache Redis cluster - get the master/slave info by following the previous answer by Abhishek.

2. If you don't have access to the cluster i.e. if you are writing an AWS Lambda or some code that is not running on the machine which can't access port 6379 of the cluster, the roundabout way is to check the IsMaster metric for each node of each shard:

    aws cloudwatch get-metric-data --cli-input-json file://test.json
   

where test.json looks like so:

{
  "MetricDataQueries": [
    {
      "Id": "is_master_test",
      "MetricStat": {
        "Metric": {
          "Namespace": "AWS/ElastiCache",
          "MetricName": "IsMaster",
          "Dimensions": [
            {
              "Name": "CacheClusterId",
              "Value": "xxxxx"
            },
             {
              "Name": "CacheNodeId",
              "Value": "0001"
            }
          ]
        },
        "Period": 60,
        "Stat": "Minimum",
        "Unit": "Count"
      },
      "Label": "is_master_test",
      "ReturnData": true
    }
  ],
  "StartTime": "2019-06-12T10:08:0000",
  "EndTime": "2019-06-12T10:09:0000"
}

If the target node is the master the output will be like so:

{
    "MetricDataResults": [
        {
            "Timestamps": [
              "2019-06-12T10:05:00Z"
            ],
            "StatusCode": "Complete",
            "Values": [
              1.0
            ],
            "Id": "is_master_test",
            "Label": "is_master_test"
        }
    ]
}

If the target node isn't the master the output will be like so:

{
    "MetricDataResults": [
        {
            "Timestamps": [
                "2019-06-12T10:05:00Z"
            ],
            "StatusCode": "Complete",
            "Values": [
                0.0
            ],
            "Id": "is_master_test",
            "Label": "is_master_test"
        }
    ]
}

Please note that I am using the metric "Minimum" to check if this node stayed as the Master for at least the last 1 minute.