Durability Tradeoffs

The appendfsync setting

AOF durability depends on how often Redis flushes the AOF buffer to disk. The appendfsync directive controls this with three modes:

always: fsync after every command — zero data loss, lowest throughput
everysec: fsync once per second — at most 1 second of data loss, good balance (recommended)
no: OS decides when to flush — highest throughput, unpredictable data loss

# Maximum durability — every write fsynced
appendfsync always

# Balance — at most 1 second of data loss (recommended)
appendfsync everysec

# Fastest — OS controls flushing, unpredictable loss
appendfsync no

What you can lose on a crash

Mode	Max data loss	Throughput
`always`	0 commands	Lowest
`everysec`	~1 second	Good
`no`	Since last OS flush	Highest
RDB only	Since last snapshot	Highest

Combining persistence with replication

Running replicas does NOT replace persistence. A replica that crashes and restarts will reload from its own persistence files (or sync from primary). Best practice: enable persistence on replicas independently, not just on the primary.

127.0.0.1:6379> CONFIG GET appendfsync
1) "appendfsync"
2) "everysec"
127.0.0.1:6379> CONFIG SET appendfsync always
OK
127.0.0.1:6379> CONFIG GET appendfsync
1) "appendfsync"
2) "always"

CONFIG GET appendfsync
CONFIG SET appendfsync always
CONFIG GET appendfsync