Skip to content
By using AWS re:Post, you agree to the AWS re:Post Terms of Use
AWS re:Postre:Post
About re:Post

Bedrock knowledge base initial sync fails with no error msg

0

I did try to create a structured knowledge base for Bedrock. Left everything to default settings and let Bedrock to create new IAM role. Things go fine, until I try to sync. Sync starts (IN_PROGRESS) for a short moment and then I get the error as shown below.

Error msg on AWS console

Most of information is for creating unstructured knowledge bases. Anyone have run to similar issue or have ideas what could be causing this? I did populate my Redshift cluster from RDS database using zero-ETL integration. Data synced fine and I can run Redshift queries. I tested syncing both ETL enabled and disabled with no diffirence.

Or if you have a link to detailed tutorial how to create a structured knowledge base for Bedrock. I've watched the re:Invent session but it didn't to deep enough into details to be helpful. https://www.youtube.com/watch?v=SdKfdkec7SQ

UPDATE:

I seem this might be region specific thing. Everything is working smoothly in eu-west-1 but above failure happens in eu-north-1. Bedrock model selection is different between regions. Could this be the root cause?

eu-north-1: eu-north-1

eu-west-1: eu-west-1

Topics
AnalyticsMachine Learning & AIGenerative AI on AWS
Tags
Knowledge Bases for Amazon BedrockAmazon Redshift
Language
English
EXPERT
asked a year ago450 views
1 Answer
0

Based on your description, it appears you're experiencing an issue with syncing your Bedrock Knowledge Base. There are several common reasons why the sync might fail without providing a clear error message:

  1. Model access permissions: The most likely cause is that you haven't subscribed to or requested access to the embedding model being used by your Knowledge Base. Even though you've created the Knowledge Base successfully, the sync process requires access to specific Bedrock models (like Amazon Titan Embed) to generate vector embeddings. Check the "Model access" section in the Bedrock console and ensure you've subscribed to the necessary models.

  2. IAM role permissions: While you mentioned letting Bedrock create a new IAM role, the default role might not have all necessary permissions. The Knowledge Base role needs permissions to call the specified Bedrock embedding model, access your Redshift data, and manage the vector store.

  3. Vector store configuration: If you're using Amazon OpenSearch Serverless as your vector store (which is often the default), ensure that the vector index is properly configured with the 'faiss' engine.

  4. Sync process limitations: There might be an ongoing sync process that's not visible in the UI, or there could be issues with the console interface itself.

To troubleshoot:

  • Check CloudTrail logs for any "StartIngestionJob" API execution details
  • Verify you have access to the embedding model being used (typically Amazon Titan Embed)
  • Refresh the page and try the sync again
  • Check if there are any size limitations or format issues with your data in Redshift

Many users have resolved similar issues by subscribing to the required embedding model in the Bedrock Marketplace, particularly the Amazon Titan Embed model.
Sources
Bedrock Data source sync does not work | AWS re:Post
Data sync failed error in Amazon Bedrock Knowledgebase | AWS re:Post
Bedrock knowledge base data source sync is not working | AWS re:Post

answered a year ago
  • Petri Kallberg EXPERT
    a year ago
    1. Titan is enabled
    2. Did try even adding Admin policy to IAM role, but no effect.
    3. This doesn't apply to structured knowledge bases.
    4. This was the initial sync failing.

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Relevant content