Hello everyone,
I’ve noticed that Pinecone takes about 2-3 minutes for an upserted vector to be available for query. My use case involves a user upserting a vector and then querying that vector immediately. A delay of 2-3 seconds would be okay, but 2-3 minutes is problematic for my use case. I am currently using the serverless option with the starter plan.
Has anyone else experienced this delay? Is this the expected behavior, or is there something I might be doing wrong?
Thanks for your help!
How many vectors do you have in your index to reach up to 2-3 minute delay? Even with a brand new index, after upserting I have to wait 3-5 seconds before the query shows the data.
This is already very bad from an automation perspective where my tests are then verifying or using the data. I either have to retry or sleep – neither of which is desirable. How are people handling this? This needs to be addressed. After upsert and return from the API, the data needs to be available.
Also see "There can be a slight delay before new or changed records are visible to queries"
I have about 6,000 vectors, which isn’t a lot. While availability is not always immediate, sometimes it is ready right away, and other times it takes 2-3 minutes before I can query it.
Yesterday, the upsertion was immediate, but today I’m frustrated because it took 2-3 minutes to upsert a batch of 20 vectors (1536 dimensions). I even upgraded to the Standard plan, which costs money, but the upsertion still takes 2-3 minutes.
Another thing I want to point out is: For example if I upsert 1 vector A, after 2-3 mins I did not see that vector available. But If I upsert another vector B, both vector A and B will be available. It is strange.
Mandatory migration to serverless should be halted until this issue is addressed. Also see my reply on the other thread.