One of our serverless clusters is not returning any data now, but was yesterday

Support
1

One of our serverless clusters stopped sending back any results this morning. We are still using pods as a backup, but we aren’t planning to do that forever. The number of vectors in the serverless cluster is the same as it was yesterday.

What can I do to investigate this?

2

Hi @michael1, when you state the index “stopped sending back any results,” do you mean that the query responses were empty?

If possible, please share the code you’re executing and a sample response (or, in the case of an error, the full stack trace of the message).

3

We are filtering on some fields that aren’t in the metadata, but we’re doing a $ne and it was working fine before.

I can reproduce it directly with the API with the following post data:

{
	"namespace": "{{ _.services.pinecone.namespace }}",
  "topK": 200,
  "includeMetadata": true,
  "filter": {
		"someColl": {"$in": ["some-id"]},
		"someValue": {"$ne": "some-value"}
	},
  "vector": [...]
}

This is failing when someValue isn’t in the metadata. I also notice that it returns docs when the someColl filter is removed, so it’s a combination of the array $in check and and the $ne