Is knn.circuit_breaker.triggered a cluster level configuration?

I am using Opensearch 2.11 on ubuntu 22.04 LTS
I configured 3 node opensearch cluster
with 32GB RAM for each node, set 16GB heap size.

I have knn index with vectors over 40m
using nmslib as hnsw engine

my index is mapped as

{
  "my_index": {
    "mappings": {
      "properties": {
        "contentName": {
          "type": "text"
        },
        "createTime": {
          "type": "long"
        },
        "contentId": {
          "type": "keyword"
        },
        "texts": {
          "type": "text"
        },
        "updateTime": {
          "type": "long"
        },
        "vector_field": {
          "type": "knn_vector",
          "dimension": 512,
          "method": {
            "engine": "nmslib",
            "space_type": "cosinesimil",
            "name": "hnsw",
            "parameters": {
              "ef_construction": 512,
              "m": 16
            }
          }
        }
      }
    }
  }
}

When I run Vector search query for this index on second node,
“knn.circuit_breaker.triggered” configuration has been changed from false to true
and second node got evicted from cluster

and After that, all indexing requests on this index has been rejected with
“Indexing knn vector fields is rejected as circuit breaker triggered.”
even though second node joined into cluster again.

My question is, Is there any way to make other opensearch node
handle indexing requests on indices mapped with knn vector field
after one Opensearch triggered circuit break on knn related operation?

Or is it simply impossible because
circuit_breaker configuration is cluster level
so I cannot isolate such problem into single node?

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.