Why is Milvus unresponsive?

The query nodes crash and any further queries that are made are not completed.

Milvus might become unresponsive if the consumed memory of the loaded collections exceeds the capacity of the cluster.

If there are multiple collections that are loaded, release some of them to free up memory. In PyMilvus, you can do it using:

   client.release_collection("collection_name")

The released collections must be loaded again if you want to search them.

If a single collection doesn't fit in the memory, you cannot query the collection to retrieve data. To recover the collection:

• Increase the query node memory by doing the following:

  • Upgrade the size.
  • Increase the query node replicas.

• Reduce the collection size by doing the following:

  • Delete rows.
  • After upgrading the size, extract the required data from the collection, drop the collection, and then reset the size. Alternatively, continue to use the upgraded size without modifying the collection.