DynamoDb bug on python boto3 query with GSI

I don't see a bug here, you are using Limit 1 and also an ExclusiveStartKey. Can you share the same code snippet with the values for your variables, that will help me determine the issue. Also, please share your Boto3 and Python version.

Here is the full code in lambda function (BOTO3 and Python 3.9)

import simplejson as json
import boto3
from boto3.dynamodb.conditions import Key

def lambda_handler(event, context):
    # TODO implement
    
    
    try:
        
        dynamodb = boto3.resource('dynamodb')
        customerID = event['pathParameters']['customer-id']
        startDate = (event['queryStringParameters']).get('start_date')
        endDate = (event['queryStringParameters']).get('end_date')
        LKCTI = (event['queryStringParameters']).get('lkcti')
        LKCI = (event['queryStringParameters']).get('lkci')
        LKTT = (event['queryStringParameters']).get('lktt')
        
        
        receivedLastEvaluatedKey = {}
        
        if LKCTI and LKCI and LKTT:
            receivedLastEvaluatedKey['coin_transaction_id'] = LKCTI
            receivedLastEvaluatedKey['customer_id'] = LKCI
            receivedLastEvaluatedKey['transaction_time'] = LKTT
            
        
            
        coinTable = dynamodb.Table('customer-coin-transaction')
        coinResp = {}
        if LKCTI and LKCI and LKTT:
            #created date is needed for sorting by created date
            coinResp = coinTable.query(
                IndexName = "customer_id-transaction_time-index",
                KeyConditionExpression = 'customer_id = :ci and transaction_time between :date1 and :date2',
                ExpressionAttributeValues = {
                    ':ci' : customerID,
                    ':date1' : startDate,
                    ':date2' : endDate,
                },
                Limit = 1,
                ExclusiveStartKey = receivedLastEvaluatedKey,
                ScanIndexForward = False,
            )
        else:
            coinResp = coinTable.query(
                IndexName = "customer_id-transaction_time-index",
                KeyConditionExpression = 'customer_id = :ci and transaction_time between :date1 and :date2',
                ExpressionAttributeValues = {
                    ':ci' : customerID,
                    ':date1' : startDate,
                    ':date2' : endDate,
                },
                Limit = 1,
                ScanIndexForward = False,
            )
        
        
        if 'Items' in coinResp:
            lastEvaluatedKey = ""
            
            if 'LastEvaluatedKey' in coinResp:
                lastEvaluatedKey = coinResp['LastEvaluatedKey']
            
            result = {
                "last_evaluated_key" : lastEvaluatedKey,
                "last_evaluated_key_old" : receivedLastEvaluatedKey,
                "items" : coinResp['Items'],
            }
            
            return {
                'statusCode': 200,
                'body': json.dumps(result)
            }
        
        else:
            return {
                'statusCode': 200,
                'body': json.dumps('')
            }
            
        return {
            'statusCode': 400,
            'body': json.dumps('failed')
        }

    except Exception as e:
        return {
            "statusCode": 400,
            "body": str(e),
        }

I call this service from a frontend, pagination only 1 row for each call with descending sorting, from first call with empty last evaluated key .

The first call result is first row with last evaluated key which is the row #1 returned.

and, when I make the second call with last evaluated key which is the row #1 (result from the first call), a result returned from the second call is row #3 with row #3 is last evalauted key returned.

and, if I make the third call with last evalauted key from second call, which is the row #3, the third call result is row #4.

Problem is the row #2 is skipped when customer_id and transaction_time is the same value as row #3. (Expected result for each call is 1->2->3->4->.... but given result for each call is 1->3->4->..... )

Row #2 will not be skipped if its transaction_time, sorting key, is different from row #3.