{
  "schema_version": 2,
  "id": "integrate/redis-data-integration/data-pipelines/rejected-records",
  "title": "Rejected records",
  "url": "https://redis.io/docs/latest/integrate/redis-data-integration/1.19.1/data-pipelines/rejected-records/",
  "summary": "Learn how RDI stores records that cannot be processed.",
  "content": "\nRedis Data Integration (RDI) sends records that it cannot process to a dead letter\nqueue (DLQ). In the Redis Cloud UI, these records are called **rejected records**.\n\nRejected records help you understand which source tables are affected and why\nspecific records could not continue through the pipeline. They are intended for\ntroubleshooting and support, not for normal pipeline operation.\n\n## When records are rejected\n\nRDI can reject a record when it cannot safely transform or write the change event.\nCommon causes include:\n\n- The incoming change event is malformed or missing required metadata.\n- A transformation job fails while processing the record.\n- A target write fails, for example because the target key already has an incompatible data type.\n\nBy default, RDI stores rejected records instead of silently dropping them. You can\nchange this behavior with the `processors.error_handling` setting. See the\n[pipeline configuration file](https://redis.io/docs/latest/integrate/redis-data-integration/1.19.1/data-pipelines/pipeline-config)\nfor more information.\n\n## How RDI stores rejected records\n\nRDI stores rejected records in the RDI database as capped Redis streams. Each DLQ\nstream corresponds to a source table and tracks the records rejected for that\ntable.\n\nDLQ stream names use the `dlq:` prefix followed by the source data stream name.\nIn current RDI versions, the stream name is typically:\n\n```text\ndlq:data:{rdi}:\u003cschema_or_database\u003e.\u003ctable\u003e\n```\n\nFor example, rejected records for the `public.users` table are stored in:\n\n```text\ndlq:data:{rdi}:public.users\n```\n\nFor sources that include the source name in the stream qualifier, the final part\ncan contain three components:\n\n```text\ndlq:data:{rdi}:\u003csource\u003e.\u003cschema_or_database\u003e.\u003ctable\u003e\n```\n\nSome RDI versions or configurations can use a hash-tagged variant such as\n`dlq:{data:rdi:\u003cschema_or_database\u003e.\u003ctable\u003e}`. To find all DLQ streams in the\nRDI database, scan for stream keys that start with `dlq:`.\n\nThe maximum number of records stored per DLQ stream is controlled by\n`processors.dlq_max_messages`. When the stream reaches the configured limit,\nolder entries are evicted as newer entries are added.\n\n## What to inspect\n\nStart with the rejected count for each affected table, then inspect a sample\nrecord from the table with the highest count or the table that matters most to\nyour application.\n\nUseful fields include:\n\n- The affected table.\n- The rejection time.\n- The rejected operation. RDI stores this as an `opcode` value such as `c`,\n  `u`, `d`, `r`, `t`, or `m`. See [Using the operation code](https://redis.io/docs/latest/integrate/redis-data-integration/1.19.1/data-pipelines/transform-examples/redis-opcode-example)\n  for the operation labels.\n- The rejection reason.\n- The transformation job or operation, when the failure happened during transformation.\n\n\u003e [!NOTE]\n\u003e Rejected records can contain source data when you inspect them directly in the\n\u003e RDI database. Treat DLQ contents as sensitive customer data. The Redis Cloud UI\n\u003e uses the RDI DLQ API and shows a sanitized set of troubleshooting metadata. It\n\u003e does not show the original record payload or every field stored in the\n\u003e corresponding DLQ stream.\n\n## Resolve rejected records\n\nUse the rejection reason to identify the likely fix:\n\n- If a transformation job failed, update the job configuration and deploy the pipeline change.\n- If target writes failed because of incompatible existing keys, update the target data or key mapping.\n- If records are malformed, inspect the source connector and source database change data capture configuration.\n\nRDI does not automatically replay records from the DLQ after you fix the cause.\nIf you need existing source data to be processed again, reset the pipeline after\napplying the fix. See [Reset data pipeline](https://redis.io/docs/latest/operate/rc/rdi/view-edit#reset-data-pipeline)\nfor Redis Cloud, or use the appropriate self-managed RDI reset workflow.\n\n## CLI and API access\n\nFor self-managed RDI, use the [`redis-di list-dlqs`](https://redis.io/docs/latest/integrate/redis-data-integration/1.19.1/reference/cli/redis-di-list-dlqs)\ncommand to see the dead-letter queues and the\n[`redis-di list-dlq-records`](https://redis.io/docs/latest/integrate/redis-data-integration/1.19.1/reference/cli/redis-di-list-dlq-records)\ncommand (also available as `redis-di get-rejected`) to inspect the rejected records of a queue.\n\nFor Redis Cloud RDI, connect to the RDI database and inspect the corresponding\nDLQ streams directly when you need details that are not shown in the Redis Cloud\nUI.\n\nRDI API v2 also includes DLQ inspection endpoints. See the\n[API reference](https://redis.io/docs/latest/integrate/redis-data-integration/1.19.1/reference/api-reference)\nfor endpoint details.\n",
  "tags": ["docs","integrate","rs","rdi"],
  "last_updated": "2026-09-19T17:55:58-07:00"
}