---
name: redshift-guide
description: "Amazon Redshift is NOT PostgreSQL; corrects LLM mistakes; covers SQL, metadata, connectivity/auth, storage, query performance, and operations. Applies ONLY to Redshift (provisioned or Serverless). Pushes back on: CREATE INDEX, string_agg, pg_catalog, text type, SERIAL, stl_query. Triggers on: Redshift SQL, Redshift CREATE TABLE, Redshift COPY/UNLOAD, slow or suddenly slow Redshift query, EXPLAIN plan, queue wait, stale statistics/ANALYZE, result cache, Redshift permission denied, Redshift disk full, storage growth, snapshot/backup cost, ghost rows/bloat, VACUUM, spill to disk, distribution skew, Redshift system views, QUALIFY, PIVOT, MERGE, materialized views, Redshift Data API (Python/boto3), Redshift workload management (WLM), concurrency scaling, Redshift resize, Spectrum, awsdatacatalog, connection failures (SSL/TLS, ODBC/JDBC, connection limit), IAM/SAML federation, IAM Identity Center, datashares. Does NOT apply to: S3 policies, Athena, Glue/Iceberg outside Redshift, Aurora, RDS, DynamoDB."
metadata:
  version: "2"
---

# Amazon Redshift Guide

## Redshift is NOT PostgreSQL (read first)

Redshift speaks PostgreSQL's wire protocol and shares much of its surface syntax, so
LLMs assume PostgreSQL behavior carries over — it frequently does not. Divergences span
system tables (`pg_catalog` is incomplete), DDL (no indexes, no sequences), functions
(`string_agg`, `SUBSTR` on tables, leader-node-only functions), types (a `text` column
becomes VARCHAR(256)), and comparison semantics (trailing blanks, unenforced constraints). **Assume
divergence and verify against the reference below — do not answer from PostgreSQL habit.**
Common PostgreSQL→Redshift divergences are in `references/redshift-sql-syntax.md`.

**Works best with** the [AWS MCP server](https://docs.aws.amazon.com/agent-toolkit/) — it runs the
AWS CLI and Redshift Data API calls below in a sandboxed, audit-logged environment. All
guidance here is plain AWS CLI and SQL and works without it.

## STEP 0: Serverless or Provisioned?

Establish this before answering — APIs, system tables, and capabilities differ. Take it
from the question when it says which one; **ask** when it does not. `SELECT version()`
does not identify it.

- **Serverless** — identified by a *workgroup* (and namespace). Data API calls take
  `--workgroup-name`; the user says "workgroup"/"Serverless".
- **Provisioned** — identified by a *cluster*. Data API calls take
  `--cluster-identifier`; the user says "cluster".

| Target | System Views | Credentials API |
|---|---|---|
| **Provisioned** | SYS_, all SVV_ + STL_, STV_, SVL_, SVCS_ (single-AZ only — disabled on Multi-AZ) | `redshift:GetClusterCredentials` |
| **Serverless** | `SYS_` + a subset of `SVV_` ONLY (no STL/STV/SVL/SVCS) | `redshift-serverless:GetCredentials` |

## Critical Facts

- **SHOW commands are the primary metadata interface** — SHOW DATABASES, SHOW SCHEMAS, SHOW TABLES, SHOW COLUMNS, SHOW TABLE, SHOW VIEW. Do NOT default to pg_catalog or information_schema. → **Load `references/redshift-sql-metadata.md` for metadata/discovery questions and any "relation does not exist" report** — it has the diagnostic flow.
- **`SYS_` views are the preferred system views** — they work everywhere. `STL_`, `STV_`, `SVL_`, and `SVCS_` are provisioned single-AZ only, and some `SVV_` views are unsupported on Serverless. → **Load `references/redshift-sql-metadata.md` for any system-view or monitoring question.**
- **`sys_load_error_detail`** for COPY debugging (not `stl_load_errors`, which is provisioned single-AZ only).
- **DATEADD/DATEDIFF** — unit-first argument order: `DATEADD(day, -30, GETDATE())`, `DATEDIFF(day, start, end)`.
- **APPROXIMATE COUNT(DISTINCT col)** — Redshift-specific, ~2% error, much faster than exact COUNT(DISTINCT) on large datasets.
- **MERGE ... REMOVE DUPLICATES** — simplified dedup when source and target have identical schemas.
- **COPY should use IAM_ROLE** (the namespace role, not the caller role) + supports MANIFEST for explicit file lists + MAXERROR for error tolerance.
- **`SUBSTR()` is leader-node-only** — works on literals but errors on table columns (`SUBSTR() function is not supported (Hint: use SUBSTRING instead)`). Use `SUBSTRING()` on columns.
- **UNIQUE / PRIMARY KEY / FOREIGN KEY are informational only** — NOT enforced (duplicate rows are accepted with no error). Optimizer hints; enforce integrity in the application or via MERGE. `NOT NULL` IS enforced.
- **`SHOW VIEW <schema.name>`** returns the definition of a regular view, materialized view, or late-binding view. MV freshness: `SVV_MV_INFO` (`is_stale`).
- **No `FILTER (WHERE ...)` on aggregates** — PostgreSQL-only, Redshift rejects it. Write `COUNT(CASE WHEN cond THEN 1 END)` / `SUM(CASE WHEN cond THEN x END)` instead. This applies to every query you emit, including ones you derive from a reference example.
- **`TOP N` and `LIMIT N` both work** (`TOP N PERCENT` does not). A `text` column becomes `VARCHAR(256)` — use `VARCHAR(max)` or explicit length.
- **Iceberg tables use `CREATE TABLE ... USING ICEBERG`** (not `STORED AS ICEBERG`, not `TABLE_FORMAT=ICEBERG`).
- **Datashares support read and write operations** — consumers can write once the producer grants write privileges. Treat "permission denied" on a datashare write as a **missing grant**, not an unsupported operation. → **Load `references/redshift-sql-metadata.md` for requirements and limits.**
- **Storage: attribute before you remediate, and never reflex-VACUUM.** Disk-full is usually NOT data growth — rule out transient query **spill** (temp blocks) and per-node/slice **skew** first, because the cluster average hides a single full node. On RA3/Serverless a high local reading is **cache**, not out-of-space, and free managed storage does not prevent a node-local Disk Full. `SVV_TABLE_INFO.unsorted` alone is misleading — gate sorting on `vacuum_sort_benefit`. **"Storage is growing / what is driving it" is a different question from "storage is high"**: it needs a trend over months (CloudWatch, not a point-in-time view), per-table attribution is always partial, and the per-table sum is never the cluster total. → **Load `references/redshift-storage-utilization.md` to attribute space, then `references/redshift-storage-remediation.md` to act.**

## Safety Guardrails

**BLOCK:** DROP DATABASE, DELETE without WHERE, publicly-accessible=true, GRANT ALL ON ALL
**WARN then confirm:** RESIZE, RESTORE, VACUUM on large tables, ALTER PASSWORD, WLM config change, TRUNCATE / DROP TABLE — **`TRUNCATE` forces an implicit COMMIT and cannot be rolled back** (it also commits other pending work in the same transaction), so name the target table and say the rows are unrecoverable before proceeding; treat it as routine only on temp/staging tables the user has already called disposable
**Confirm:** CREATE, GRANT specific, COPY, UNLOAD

## Security Considerations

Apply these defaults when generating anything that connects, loads, or exports. Details
are in the reference files noted.

- **In transit:** the Data API is HTTPS-only. For JDBC/ODBC set the `require_ssl`
  parameter and connect with `sslmode=verify-full` so the server certificate is checked.
- **At rest:** keep cluster/namespace encryption enabled, and add
  `ENCRYPTED KMS_KEY_ID '<arn>'` to `UNLOAD` — it writes query results to S3, outside
  Redshift's own encryption. → `references/redshift-sql-ddl-copy.md`
- **Credentials:** prefer `SecretArn` (Secrets Manager) or IAM Identity Center; `DbUser`
  is acceptable because it issues temporary credentials. Never place database passwords in
  code, environment variables, or SQL text. → `references/redshift-sql-recipes-load-api.md`
- **Least privilege:** scope the namespace `IAM_ROLE` to the specific bucket and prefix
  (`s3:GetObject` on `arn:aws:s3:::<bucket>/<prefix>/*`), not `s3:*` or a managed
  full-access policy, and condition its trust policy on both `aws:SourceArn` (the
  cluster/namespace ARN) and `aws:SourceAccount` — `SourceArn` alone still allows another
  resource in the account to assume it. Grant per-object privileges rather than
  `GRANT ALL ON ALL`.
- **Audit:** CloudTrail records `redshift-data:*` API calls but not the SQL executed;
  enable Redshift audit logging (`useractivitylog`, `connectionlog`, `userlog`) for that.
  Both capture query text and user activity, so encrypt every destination in use:
  the CloudWatch Logs group (`aws logs associate-kms-key`), the CloudTrail trail
  (SSE-KMS), and the audit-log S3 bucket (SSE-S3 — audit logging to S3 supports only
  S3-managed keys, not KMS). Serverless only supports sending audit logs to CloudWatch.
- **Network:** keep `PubliclyAccessible=false` and connect over a VPC
  endpoint. Do not open port 5439 to `0.0.0.0/0` or `::/0` — scope inbound rules to
  specific CIDRs or to a referencing security group.
- **Sensitive data:** Data API results persist for 24h and `sys_load_error_detail` can
  echo fragments of rejected rows, so treat statement IDs and load-error output as
  sensitive.
- **Further reading:**
  [Security in Amazon Redshift](https://docs.aws.amazon.com/redshift/latest/dg/c_security-overview.html)
  for the full guidance behind these defaults.

## Routing Table

**MANDATORY:** When a question matches a row below, you MUST load and read the referenced file BEFORE answering.

**Ask whether the target is provisioned or Serverless before giving troubleshooting steps —
unless the question already says which one, in which case use that and do not re-confirm.**

| User Intent | Route To |
|---|---|
| "CREATE TABLE", "DISTKEY/SORTKEY", "ENCODE", "IDENTITY", "COPY", "UNLOAD", "IAM_ROLE", "Iceberg table" | `references/redshift-sql-ddl-copy.md` |
| "LISTAGG", "DATEADD/DATEDIFF", "NVL/DECODE", "type mapping", "text type", "VARBYTE", "recursive CTE" | `references/redshift-sql-functions-types.md` |
| "QUALIFY", "PIVOT/UNPIVOT", "MERGE", "TOP N", "SUBSTR error", "UNIQUE/PK not enforced", "trailing blanks", "leader-node function", "JSON", "SUPER", "PartiQL", "nested/semi-structured data" | `references/redshift-sql-extensions-semantics.md` |
| "system view", "SVV_/SYS_", "SHOW commands", "STL vs SYS", "list tables", "distkey/sortkey lookup", "datashare discovery", "2-part vs 3-part", "permission denied", "GRANT", "privileges", **"relation/table does not exist"** | `references/redshift-sql-metadata.md` |
| "how do I write SQL", "PostgreSQL vs Redshift", "which SQL reference", general dialect question | `references/redshift-sql-syntax.md` (index of the 6 SQL references + PostgreSQL-vs-Redshift failure table) |
| "COPY failed", "load error", "Data API poll", "async query", "Data API throttle" | `references/redshift-sql-recipes-load-api.md` |
| "materialized view", "MV refresh", "AUTO REFRESH", "stale view" | `references/redshift-sql-materialized-views.md` |
| "disk full" / error 1016 / 34507, "PercentageDiskSpaceUsed", "storage grew", "what's consuming space", "ghost rows", "tombstone", "spill to disk", "managed storage vs local SSD", "datashare cache", "one node full / storage skew", "Serverless storage inflation", "what is driving the storage increase" / "storage keeps climbing" / "grew N% per month" / "which tables are growing" / "is vacuum missing" — a growth-over-time question, answered by the *"Is it growing, and what is driving it?"* section: split it into four (is it really that full, is it vacuum debt, how much would VACUUM return, what is driving the trend), trend from CloudWatch over months, per-table attribution from `SYS_VACUUM_HISTORY`, and state plainly what cannot be attributed from customer-side views, any snapshot / backup / recovery-point billing or cost question — including "AWS Backup", "cross-region snapshot copy", "backup billed at full size", "why is snapshot storage this expensive", "incremental so it should be cheaper" (answer from the per-type × per-platform matrix in S16, never from AWS Backup's own incremental-copy documentation, which describes AWS Backup's behavior and not Redshift's manual-snapshot metering), "cluster unavailable / stuck after hitting 100% storage" (confirm the fill first — `PercentageDiskSpaceUsed` in CloudWatch, since the SQL views are unreachable while the cluster is down, and rule out the non-storage causes before assuming disk) — attribute the space first | `references/redshift-storage-utilization.md` |
| "VACUUM won't finish / times out", "which VACUUM mode — DELETE ONLY / SORT ONLY / FULL / RECLUSTER", "vacuum_sort_benefit", "unsorted", "reclaim space", "deep copy", "auto-vacuum CPU", "distribution key skew", "recovering a cluster stuck or unavailable at ~100% storage — headroom plus escalation", "resize failed and stuck in canceling-resize" — take action | `references/redshift-storage-remediation.md` |
| "slow Redshift query", "query was fast now slow", "why is my query slow", "query performance", "time decomposition", "queue vs execution time", "planning/compile time", "result cache hit / false fast", "EXPLAIN plan", "DS_BCAST_INNER / broadcast / nested loop / merge join", "stale statistics / ANALYZE / bad row estimate", "function on join key / plan flip", "distribution skew / skew_rows / one slice", "did the query change / query hash", "row multiplication / join fan-out / DISTINCT slow", "sys_query_history / sys_query_detail" | `references/redshift-sql-query-performance.md` |
| "ready-to-run performance query", "which SQL to run", "diagnostic queries", "execution step / skew / spill", "plan diff between runs", "svv_table_info table health", "auto-analyze / auto-vacuum history", "lock wait / transaction history" | `references/redshift-sql-query-performance-queries.md` |
| "connection timed out/refused", "SSO/federated login fails", "Response signature invalid", "assumeRoleWithSAML", "RedshiftDbUser principal tag", "sts:TagSession AccessDenied", "IAM Identity Center region", "connection limit exceeded", "ODBC/JDBC driver", "ODBC 1.x EOL/CVE", "SSL error/certificate verify failed", "permission denied to query database created from datashare", "SYSLOG ACCESS/system-table row visibility", "GetClusterCredentials AccessDenied" | `references/redshift-auth-authorization.md` |
| General Redshift question not matching above | Answer directly from general knowledge |
| Aurora, RDS, DynamoDB, Athena (non-Redshift) | **REFUSE.** State this skill is for Amazon Redshift only. Do not provide guidance for other database services. |

## Data API Quick Reference

→ **Load `references/redshift-sql-recipes-load-api.md` before answering ANY Data API, COPY-error, or async-query question.** It carries the bounded poll loop, the `HasResultSet` and `ResourceNotFoundException` handling, the per-target parameters, and the auth options.

Data API calls are **async by default** — use long polling (`--wait-time-seconds`, 1–30)
rather than blind sleeps, and keep a bounded loop for work that can exceed 30s.
Serverless takes `--workgroup-name`, provisioned takes `--cluster-identifier`.