AOF Persistence

Introduction

Append-Only File (AOF) logs every write command — on restart Redis replays the log to rebuild state. appendfsync everysec (default) balances durability (~1s loss window) with throughput; always fsyncs every write but kills perf.

BGREWRITEAOF compacts AOF by rewriting current memory state — without rewrite the file grows forever. Hybrid RDB+AOF preamble in Redis 7 speeds recovery.

Choose AOF when losing last minute of RDB snapshot is unacceptable.

Understanding the topic

Key concepts

• appendonly yes — enable AOF.
• appendfsync always | everysec | no.
• AOF rewrite — BGREWRITEAOF background compact.
• auto-aof-rewrite-percentage triggers rewrite.
• redis-check-aof validates file before start.
• Multi-part AOF (Redis 7) for faster recovery.

flowchart LR
  Writes --> Memory
  Memory -->|fork| RDB
  Memory -->|append| AOF

Step-by-step explanation

1. Write command succeeds in memory first.
2. Command appended to AOF buffer.
3. fsync policy flushes buffer to disk.
4. Rewrite fork child writes minimal new AOF.
5. Startup replays AOF commands in order.

Syntax reference

Common commands

• everysec — industry default.
• Rewrite during low traffic.
• Monitor aof_current_size growth.

CONFIG SET appendonly yes
CONFIG GET appendfsync
BGREWRITEAOF
INFO persistence

Informative example

Enable AOF on staging to match production durability:

redis-cli CONFIG SET appendonly yes
redis-cli CONFIG SET appendfsync everysec
redis-cli CONFIG REWRITE
redis-cli INFO persistence | grep aof

CONFIG REWRITE persists to redis.conf if writable. Managed Redis sets this in console — don't fight provider defaults.

Real-world use

Real-world use cases

• Session store needing <1 min RPO.
• Primary ephemeral DB with replay recovery.
• Audit-sensitive write log (supplement not replace audit DB).
• Pair RDB backup + AOF live durability.
• Regulatory require durable write trail.

Best practices

• appendfsync everysec default for most apps.
• Monitor rewrite duration and failures.
• Disk SSD for AOF — HDD struggles on always.
• auto-aof-rewrite-min-size avoid tiny rewrites.
• Test full AOF reload in staging.
• Hybrid snapshot preamble when available.

Common mistakes

• appendfsync no — OS cache only, big loss on crash.
• Never rewriting — disk fills.
• AOF on same disk as logs without IOPS headroom.
• Assuming AOF replaces off-site RDB backups.

Advanced interview questions

Summary

AOF logs writes — finer RPO than RDB alone.