Redis Study Notes in Mind Map Form

These notes organize Redis into a structured set of mind maps for systematic study, covering core data structures, typical use cases, recurring problems, performance tuning, persistence, cluster modes, submodules, and a brief look at the kernel startup path.

The main areas include:

• core data types and where they fit
• analysis of common issues
• performance optimization
• persistence
• cluster modes
• submodules

Core concepts

Data types and typical scenarios

Basic data types

• string
• Binary-safe, so it can store any kind of data.

• A single key can hold up to 512 MB.

• hash

• A collection of field-value pairs.

• Well suited for storing, reading, and updating user attributes.

• list

• Implemented as a linked list (specifically a doubly linked list).

• Useful for features such as latest-message feeds, for example a social timeline, and also for message queues.

• set

• Implemented using a hash table, with no duplicate elements.
• Good for cases that depend on uniqueness, such as counting all distinct IPs that visited a site.

• Also useful in friend recommendation: intersect tags, and if the overlap exceeds a threshold, a recommendation can be made.

• sorted sets

• Add a score to each Set element, and elements are ordered by that score.

• Commonly used for leaderboards and weighted message queues.

• bitmaps

• Store 0 or 1 as bit values in a map-like structure.

• Can be used for user check-ins, online-status statistics across millions of users, and deduplication across tens of millions of consumers.

• HyperLogLogs

• Accept multiple input elements and return an estimated cardinality.

Common use cases

Redis is often used for:

• caching data
• latest-message ranking and feed-style features such as social timelines
• message queues and weighted message queues
• finding mutual friends
• friend recommendations based on tag intersections that exceed a threshold
• leaderboards
• user check-ins
• online-status statistics for millions of users
• deduplication across tens of millions of consumers
• distributed locks

Common issue analysis

Client-side errors

Could not get a resource from the pool

Possible causes include:

• connection leakage, where borrowed connections are not returned to the pool
• concurrency that is too high for the configured maximum pool size
• slow commands being executed

Unexpected end of stream

Possible causes include:

• a Jedis object or a pipeline object being accessed by multiple threads
• the client buffer being full

Redis uses different client buffers:

• normal client buffer: used for ordinary commands such as get, set, mset, hgetall, zrange, and similar operations
• slave client buffer: used to receive write commands from the master node during replication
• pub/sub buffer: pub/sub is not treated as ordinary commands, so it uses a separate buffer

ERR illegal address

• the configured IP address or port is incorrect

ERR max number of clients reached

Possible causes include:

• the Redis instance has exceeded its configured maxclients
• the server-side maxclient setting is too small
• too many client connection pools, or pools that are too large
• client-side connection leakage, while the server does not close idle connections on a schedule

java.net.SocketTimeoutException: Read timed out

Possible causes include:

• read/write timeout values are set too aggressively
• slow queries or Redis-side blocking
• unstable network conditions

NOAUTH Authentication required

• the client did not send a password

OOM command not allowed when used memory > 'maxmemory'

• the Redis instance has run out of writable memory
• increase maxmemory; values larger than 10 GB are not recommended

LOADING Redis is loading the dataset in memory

• when Redis is loading a persistence file into memory, normal read/write operations are not available

java.net.SocketTimeoutException: connect timed out

Possible causes include:

• connection timeout is set too short
• tcp-backlog is full, causing new connections to fail
• network connectivity between client and server is abnormal

Please close pipeline or multi block before calling this method.

• use pipeline.syncAndReturnAll() to handle it

Server-side errors

psync scheduled to be closed ASAP for overcoming of output buffer limits

• set client-output-buffer-limit slave to client-output-buffer-limit slave 0 0 0

Performance optimization

Mind map:

Persistence

Mind map:

Cluster modes

Mind map:

Submodules

Mind map:

Redis internals

The kernel-level section currently includes:

• the kernel startup process

Kernel startup process

The kernel portion can be expanded further into more detailed notes based on the material in the mind maps.