3.4 Redis Blocking · ThinkerQAQ

This is a historical learning note and may contain outdated or incomplete understanding.

1. What Is Redis Blocking?

In essence, Redis has a single-threaded architecture and command execution is completed in this thread.

A business-side Redis client times out when connecting to the Redis server, triggering a monitoring alert.

Unreasonable data structures
- How to discover:
  - Check the slow-query log: Redis Slow Queries
- How to solve:
  - Redis bigkey
  - Redis hotkey
Too much data in an atomic command, for example mset with 10K pieces of data.
- How to solve: execute in batches; if atomicity is not required, pipeline can be used.

CPU saturation means Redis drives a single CPU core close to 100% utilization.
How to discover:
- Use the top command to easily identify the CPU utilization of the corresponding Redis process.
How to solve:
- In essence, this is a single-instance Redis performance bottleneck, so Redis Cluster mode is needed to distribute the pressure.

Some persistence operations can block the main thread: fork blocking and AOF flush blocking.
Fork blocking
- RDB and AOF rewriting fork a child process for processing.
- How to discover: use info stats and check latest_fork_usec.
- How to solve: avoid excessively large memory instances.
AOF flush blocking
- Redis Persistence

For example, Redis is deployed together with other services on the same machine.

Set /proc/sys/vm/swapness to 1, retaining swap while reducing its use.

Network interruption: for example, network switching or bandwidth exhaustion; sar can be used for diagnosis.
Redis maximum number of client connections: Redis configuration file.
Connection overflow: for example, the Linux process file-descriptor limit.

Depends on the network environment from client -> server.