quotas are exceeded error while deploying Vertex AI Matching Engine Index

Question

I am following this example notebook and used custom word embeddings of my own.

i am getting the following error while deploying the index at this step

r = index_endpoint_client.deploy_index(
    index_endpoint=INDEX_ENDPOINT_NAME, deployed_index=deploy_ann_index
)

ResourceExhausted: 429 The following quotas are exceeded: MatchingEngineDeployedIndexNodes

i have list endpoint deployed that was done in previous step.

ListIndexEndpointsPager<index_endpoints {
  name: "projects/xxxxxxx/locations/us-central1/indexEndpoints/xxxxxxxx"
  display_name: "index_endpoint_for_demo"
  etag: "AMEw9yPT2hoRkYEEApekieucfedTYyEb5prjgG60WL1pVUfxxxxxxxxxxx"
  create_time {
    seconds: 1645125657
    nanos: 225990000
  }
  update_time {
    seconds: 1645125657
    nanos: 794785000
  }
  network: "projects/xxxxx/global/networks/ucaip-haystack-vpc-network"
}

Answer 1

If an index is using multiple shards, you need to calculate your available quota by dividing it by number of shards used for 1 replica count. For example, if your index is spread across 10 shards and you have quota of 50, at max you can set max_replica_count to 5 (50/10).

You can check how many shards your index uses by going to Metrics Explorer and looking up Matching Engine metrics, see more information here.