AI-200 practice: Implement a change feed processor to detect and handle new or updated items

Question 1

You need multiple instances of a serverless function to process the Cosmos DB change feed for a container in parallel, with automatic load balancing across instances and the ability to resume from where processing left off after a restart.

What must you configure?

Accepted Answer

A lease container that the change feed processor uses for checkpoints so parallel instances can divide partitions and resume after restarts

Answer

An in-memory polling loop that queries documents by _ts on a timer and stores the last processed timestamp only inside each function instance

Answer

A Service Bus queue that receives every Cosmos DB write via a stored trigger, replacing the change feed processor with queue-based fan-out

Answer

A pre-trigger stored procedure that runs on each write and synchronously copies the changed document into Blob Storage for later batch processing

Question 2

You need your application to react in near-real time whenever new documents (containing chatbot transcripts) are inserted into a Cosmos DB for NoSQL container, so they can be embedded and indexed for search. What should you implement?

Accepted Answer

Change feed processor on the container

Answer

Nightly scheduled batch export job run

Answer

Hourly polling loop querying the container

Answer

Manually triggered Cosmos stored procedure

Question 3

A Cosmos DB for NoSQL change feed processor for an AI indexing pipeline occasionally reprocesses the same documents after a host restart. What is the most likely cause and fix?

Accepted Answer

Lease checkpoints lagged; tune lease checkpointing

Answer

Change feed is at-most-once; Cosmos has a bug

Answer

Partition key design alone caused reprocessing

Answer

Lower consistency to Eventual to stop duplicates

Question 4

In Azure Cosmos DB for NoSQL, what mechanism allows an application to detect and react to newly inserted or updated items in near real time?

Accepted Answer

Change feed stream

Answer

Time to Live (TTL)

Answer

Partition key range

Answer

Indexing policy set

Question 5

You are using the Azure Cosmos DB Change Feed Processor to update a vector index in Azure AI Search. What is the role of the 'Lease Container' in this architecture?

Accepted Answer

Coordinate checkpoints and worker lease ownership

Answer

Persist completed vector search result document sets

Answer

Cache inbound user search request payload documents

Answer

Store backup copies of embedding vector field values

Question 6

In the Azure Cosmos DB Python SDK, you want to use the Change Feed. Which mode allows you to see all historical versions of a document, including intermediate updates and deletes?

Accepted Answer

All versions and deletes

Answer

Latest version only mode

Answer

Full fidelity change mode

Answer

Incremental snapshot mode

Question 7

A change feed processor named `embedder` monitors a container whose feed is split into 4 leases. To speed up processing, you start 6 instances of the processor, all with the same processor name and lease container. What happens?

Accepted Answer

The 4 leases are spread across the instances, so at most 4 do work and 2 sit idle until a lease becomes available

Answer

Each change is delivered to all 6 instances, so every document is embedded 6 times

Answer

The processor fails to start because the instance count must equal the lease count

Answer

Changes are split round-robin across all 6 instances, so each instance gets about a sixth of the work

Question 8

A change feed processor in latest version mode keeps a vector search index in sync with a Cosmos DB container. When source documents are deleted, their vectors stay in the search index. What is the standard way to make deletes visible to this processor?

Accepted Answer

Mark documents as deleted (for example isDeleted: true), let the processor remove them from the index, and set a TTL on the item so Cosmos DB removes it later

Answer

Restart the processor with a start time of the beginning of time so that it replays deletions

Answer

Raise the lease renew interval so that the processor has time to observe delete operations

Answer

Switch the account to Strong consistency, which makes deletes appear in the change feed

Question 9

You deploy a new .NET change feed processor to backfill embeddings for the 2 million documents already in a container, then keep processing new changes. With a fresh, empty lease container, the processor handles only documents written after it starts. What should you add to the builder before the first start?

Accepted Answer

.WithStartTime(DateTime.MinValue.ToUniversalTime())

Answer

.WithPollInterval(TimeSpan.Zero)

Answer

.WithMaxItems(10000)

Answer

.WithInstanceName("backfill")

Question 10

Which two statements about the lease container used by the Cosmos DB change feed processor are true? Select two.

Accepted Answer

The lease container must use /id as its partition key

Accepted Answer

Processors with different processor names can share one lease container, and each keeps its own independent leases

Answer

The change feed processor guarantees exactly-once delivery of each change

Answer

The lease container must be in a different Cosmos DB account from the monitored container

Answer

The change feed guarantees a single global order of changes across all partition key values

11 questions