|
| 1 | +# How to setup DuckDB node in AliSQL |
| 2 | + |
| 3 | +[ [快速部署 AliSQL DuckDB 节点](./how-to-setup-duckdb-node-zh.md) | [How to setup DuckDB node in AliSQL](./how-to-setup-duckdb-node-en.md) ] |
| 4 | + |
| 5 | +## Overview |
| 6 | + |
| 7 | +AliSQL integrates DuckDB as an analytical storage engine. |
| 8 | + |
| 9 | +- **User data** is stored in **DuckDB** |
| 10 | +- **System tables and metadata** remain in **InnoDB** (e.g., `mysql.*`, data dictionary) |
| 11 | + |
| 12 | +You can: |
| 13 | +1) bootstrap a **brand-new instance** where new tables default to DuckDB |
| 14 | +2) **convert an existing InnoDB instance** to DuckDB at startup |
| 15 | +3) build a **DuckDB replica (HTAP)** to replay binlogs from an OLTP primary |
| 16 | + |
| 17 | +--- |
| 18 | + |
| 19 | +## 1. Bootstrap a Brand-New DuckDB Instance |
| 20 | + |
| 21 | +Use this when you are creating a fresh instance and want newly created tables to land in DuckDB by default. |
| 22 | + |
| 23 | +### 1.1 Install AliSQL 8.0.44 |
| 24 | + |
| 25 | +#### Option A: Build from source |
| 26 | +```bash |
| 27 | +git clone https://github.com/alibaba/AliSQL.git |
| 28 | +cd AliSQL |
| 29 | + |
| 30 | +sh build.sh -t release -d /opt/alisql |
| 31 | +make install |
| 32 | +``` |
| 33 | + |
| 34 | +#### Option B: Install from RPM |
| 35 | +```bash |
| 36 | +# For example, use x86 el6 rpm for installation |
| 37 | +wget https://github.com/alibaba/AliSQL/releases/download/AliSQL-8.0.44-1/alisql-8.0.44-1.el8.x86_64.rpm |
| 38 | +rpm -ivh alisql-8.0.44-1.el8.x86_64.rpm |
| 39 | +``` |
| 40 | + |
| 41 | +--- |
| 42 | + |
| 43 | +### 1.2 Directory bootstrap + config generator script |
| 44 | + |
| 45 | +```bash |
| 46 | +cat > init_alisql_dir.sh <<'EOF' |
| 47 | +#!/usr/bin/env bash |
| 48 | +set -euo pipefail |
| 49 | +
|
| 50 | +# Usage: |
| 51 | +# ./init_alisql_dir.sh <absolute_path> |
| 52 | +
|
| 53 | +if [[ $# -ne 1 ]]; then |
| 54 | + echo "Usage: $0 <absolute_path>" |
| 55 | + exit 1 |
| 56 | +fi |
| 57 | +
|
| 58 | +DIR="$1" |
| 59 | +if [[ "$DIR" != /* ]]; then |
| 60 | + echo "Error: Argument must be an ABSOLUTE PATH (starting with /)." |
| 61 | + exit 1 |
| 62 | +fi |
| 63 | +
|
| 64 | +DIR="${DIR%/}" |
| 65 | +
|
| 66 | +mkdir -p \ |
| 67 | + "$DIR/data/dbs" \ |
| 68 | + "$DIR/data/mysql" \ |
| 69 | + "$DIR/run" \ |
| 70 | + "$DIR/log/mysql" \ |
| 71 | + "$DIR/tmp" |
| 72 | +
|
| 73 | +cat > "$DIR/alisql.cnf" <<EOF2 |
| 74 | +[mysqld] |
| 75 | +# --- Paths --- |
| 76 | +basedir = |
| 77 | +datadir = ${DIR}/data/dbs |
| 78 | +socket = ${DIR}/run/mysql.sock |
| 79 | +pid-file = ${DIR}/run/mysqld.pid |
| 80 | +tmpdir = ${DIR}/tmp |
| 81 | +
|
| 82 | +# --- Basic --- |
| 83 | +character_set_server = utf8mb4 |
| 84 | +skip_ssl |
| 85 | +lower_case_table_names = 1 |
| 86 | +core-file |
| 87 | +
|
| 88 | +# --- Logs --- |
| 89 | +log-error = ${DIR}/log/mysql/error.log |
| 90 | +
|
| 91 | +# Binary logging (optional) |
| 92 | +log-bin = ${DIR}/log/mysql/mysql-bin |
| 93 | +binlog_format = ROW |
| 94 | +
|
| 95 | +# --- InnoDB (system tables / metadata) --- |
| 96 | +innodb_data_home_dir = ${DIR}/data/mysql |
| 97 | +innodb_log_group_home_dir = ${DIR}/data/mysql |
| 98 | +
|
| 99 | +# --- DuckDB Storage Engine --- |
| 100 | +duckdb_mode=ON |
| 101 | +force_innodb_to_duckdb=ON |
| 102 | +
|
| 103 | +# DuckDB resources (0 = auto) |
| 104 | +duckdb_memory_limit=0 |
| 105 | +duckdb_threads=0 |
| 106 | +duckdb_temp_directory=${DIR}/tmp |
| 107 | +EOF2 |
| 108 | +
|
| 109 | +echo "OK: created $DIR and $DIR/alisql.cnf" |
| 110 | +EOF |
| 111 | + |
| 112 | +chmod +x init_alisql_dir.sh |
| 113 | +``` |
| 114 | + |
| 115 | +References: |
| 116 | +- MySQL 8.0 manual: https://dev.mysql.com/doc/refman/8.0/en/ |
| 117 | +- DuckDB variables: [AliSQL DuckDB variables](./duckdb/duckdb_variables-zh.md) |
| 118 | + |
| 119 | +--- |
| 120 | + |
| 121 | +### 1.3 Initialize and start |
| 122 | + |
| 123 | +```bash |
| 124 | +# Example: store all data under $HOME/alisql_8044 |
| 125 | +./init_alisql_dir.sh $HOME/alisql_8044 |
| 126 | + |
| 127 | +# Initialize |
| 128 | +/opt/alisql/bin/mysqld --defaults-file=$HOME/alisql_8044/alisql.cnf --initialize-insecure |
| 129 | + |
| 130 | +# Start, the port is 3306 by default |
| 131 | +/opt/alisql/bin/mysqld_safe --defaults-file=$HOME/alisql_8044/alisql.cnf & |
| 132 | +``` |
| 133 | + |
| 134 | +--- |
| 135 | + |
| 136 | +### 1.4 Validate and use |
| 137 | + |
| 138 | +```sql |
| 139 | +CREATE DATABASE test; |
| 140 | +USE test; |
| 141 | + |
| 142 | +-- Create DuckDB table |
| 143 | +CREATE TABLE t ( |
| 144 | + id INT PRIMARY KEY, |
| 145 | + name VARCHAR(255) |
| 146 | +) ENGINE = DuckDB; |
| 147 | + |
| 148 | +-- Convert |
| 149 | +ALTER TABLE t ENGINE = InnoDB; |
| 150 | +ALTER TABLE t ENGINE = DuckDB; |
| 151 | + |
| 152 | +-- DML |
| 153 | +INSERT INTO t VALUES (1, 'John'); |
| 154 | +SELECT * FROM t; |
| 155 | +UPDATE t SET name = 'Jane' WHERE id = 1; |
| 156 | +DELETE FROM t WHERE id = 1; |
| 157 | + |
| 158 | +-- DDL |
| 159 | +ALTER TABLE t ADD COLUMN age INT; |
| 160 | +ALTER TABLE t RENAME TO t_new; |
| 161 | +``` |
| 162 | + |
| 163 | +--- |
| 164 | + |
| 165 | +## 2. One-Click Conversion from an Existing InnoDB Instance |
| 166 | + |
| 167 | +Use this when you want to migrate existing InnoDB tables to DuckDB to boost analytical performance. |
| 168 | + |
| 169 | +### Steps |
| 170 | + |
| 171 | +1) **Stop the existing instance** (clean shutdown) |
| 172 | + |
| 173 | +2) **Update `my.cnf`** |
| 174 | +```ini |
| 175 | +[mysqld] |
| 176 | +duckdb_mode=ON |
| 177 | +force_innodb_to_duckdb=ON |
| 178 | + |
| 179 | +duckdb_memory_limit=0 |
| 180 | +duckdb_threads=0 |
| 181 | +duckdb_temp_directory=/path/to/duckdb_temp_dir |
| 182 | + |
| 183 | +duckdb_convert_all_at_startup=ON |
| 184 | +duckdb_convert_all_at_startup_threads=32 |
| 185 | +duckdb_convert_all_at_startup_ignore_error=ON |
| 186 | +``` |
| 187 | + |
| 188 | +3) **Start the instance** |
| 189 | + |
| 190 | +4) **Verify** |
| 191 | +```sql |
| 192 | +-- Check the conversion stage |
| 193 | +SHOW GLOBAL STATUS LIKE 'DuckDB_convert_stage_at_startup'; |
| 194 | + |
| 195 | +-- Check the converted tables |
| 196 | +SELECT table_schema, table_name, engine |
| 197 | +FROM information_schema.tables |
| 198 | +WHERE engine = 'DuckDB'; |
| 199 | +``` |
| 200 | + |
| 201 | +### Notes |
| 202 | + |
| 203 | +- Startup conversion is resource-intensive (CPU/IO). Monitor the error log for progress and failures. |
| 204 | +- Ensure enough free disk space (recommended: ≥ original data size). |
| 205 | +- Backup before conversion (backup set + cnf). |
| 206 | +- Upgrade MySQL 5.7 or below to 8.0 first. |
| 207 | +- If your source instance is not 8.0.44, an upgrade may be triggered on startup; ensure a clean shutdown beforehand. |
| 208 | + |
| 209 | +--- |
| 210 | + |
| 211 | +## 3. DuckDB Replica (HTAP) |
| 212 | + |
| 213 | +A common pattern is OLTP on the primary and OLAP on a DuckDB replica, fed by row-based binlogs. |
| 214 | + |
| 215 | +### Steps |
| 216 | + |
| 217 | +1) **Prepare the primary** |
| 218 | +- enable binlog |
| 219 | +- `binlog_format=ROW` |
| 220 | +- enable GTID |
| 221 | + |
| 222 | +1) **Configure the replica `my.cnf`** |
| 223 | +```ini |
| 224 | +[mysqld] |
| 225 | +duckdb_mode=ON |
| 226 | +duckdb_require_primary_key=ON |
| 227 | +force_innodb_to_duckdb=ON |
| 228 | + |
| 229 | +dml_in_batch=ON |
| 230 | +update_modified_column_only=ON |
| 231 | +duckdb_multi_trx_in_batch=ON |
| 232 | +duckdb_multi_trx_timeout=5000 |
| 233 | +duckdb_multi_trx_max_batch_length=268435456 |
| 234 | +``` |
| 235 | + |
| 236 | +3) **Initialize the replica** |
| 237 | +Follow section 1 (new instance) or section 2 (optional startup conversion), depending on your scenario. |
| 238 | + |
| 239 | +4) **Set up replication** |
| 240 | +```sql |
| 241 | +CHANGE MASTER TO |
| 242 | + MASTER_HOST='your-master-ip', |
| 243 | + MASTER_USER='repl', |
| 244 | + MASTER_PASSWORD='your-password', |
| 245 | + MASTER_AUTO_POSITION=1; |
| 246 | + |
| 247 | +START SLAVE; |
| 248 | +``` |
| 249 | + |
| 250 | +5) **Done** |
| 251 | +Route analytics/reporting queries to the DuckDB replica. |
0 commit comments