Step 8: Validate in three levels

Validate the collection in increasing depth before you trust it with production traffic.

Level 1: The collection is reachable

Confirm the Migration Console can reach and authenticate to the collection:

console clusters connection-check --cluster target
console clusters curl target /

Level 2: Index and document counts match

Compare index lists and per-index document counts between the source domain and the collection:

console clusters cat-indices --refresh
console clusters curl target /<index>/_count

Document counts on the target may differ slightly from the source if the source had deleted-but-not-merged documents; investigate only material gaps. Use the failed-document Amazon CloudWatch Logs Insights query in Verifying no failed documents to confirm nothing was dropped.

Level 3: Representative vector queries

For a vector-search workload, run representative k-NN queries against the collection and confirm the results match what the source returns for the same query vectors:

console clusters curl target /<index>/_search?pretty -XPOST \
  -H 'Content-Type: application/json' \
  -d '{
    "size": 5,
    "query": {
      "knn": {
        "<vector-field>": { "vector": [0.1, 0.2, 0.3], "k": 5 }
      }
    }
  }'

Verify that the dense_vector-to-knn_vector field-type transformation produced the expected mapping on the target, and that scores and ranking are consistent with the source for a sample of known queries.