Safely Accessing Massive Redis Data Without Causing Service Outage

The author shares how to access massive data in Redis while avoiding incidents.

Incident: Using the KEYS user_token* command to count online users caused Redis to become unresponsive because KEYS scans the entire keyspace with O(N) complexity on a single‑threaded server.

Analysis: With millions of keys, KEYS blocks the server; Redis processes commands sequentially, so other operations must wait for KEYS to finish.

Solution: Use the SCAN command, which iterates incrementally using a cursor, does not block the thread, supports pattern matching and a COUNT hint, may return duplicate entries, and requires client‑side deduplication; the iteration ends when the returned cursor is zero.

SCAN command format: SCAN cursor [MATCH pattern] [COUNT count]

Explanation: SCAN is incremental, returning a small subset of elements each call; the cursor starts at 0 and the iteration finishes when the cursor returns to 0, preventing Redis from hanging.

Example: Starting from cursor 0 returns cursor 6 and some keys; subsequent SCAN calls continue from cursor 6, and so on, until the cursor reaches 0.

Conclusion: SCAN is a common interview topic and essential for handling large keyspaces in production; using KEYS on large datasets can severely degrade performance.