This is a transient startup message: Redis restarted and is reading its persistence file back into memory, so it cannot serve commands yet. Waiting and retrying resolves it. Jump to your situation below or work through the methods in order.
The LOADING Redis is loading the dataset in memory message is transient and expected: Redis has just restarted and is reading its RDB snapshot or AOF file from disk back into memory before it can serve commands. Until that finishes, most commands are rejected with this message. For a large dataset the load can take a while. The fix is mostly to wait and to make your client resilient: add a retry with exponential backoff so transient LOADING (and similar) responses are retried rather than crashing the app. If loading is unusually slow or never completes, look at the size of the persistence file, disk speed, and whether an oversized AOF could be shrunk with a rewrite. This is not corruption; it is normal warm-up.
Redis serves from memory, so when it starts it must first load the dataset from its persistence file, an RDB snapshot or the AOF log, into RAM. While that is happening it is not ready to answer queries, so it returns LOADING for most commands.
That makes this a timing condition, not an error in your data or code. The cure is patience plus a client that expects it: retrying with backoff rides out the warm-up. Persistent slowness points at a large file or slow disk rather than a fault.
“LOADING is not a problem, it is Redis warming up. It restarted and it is pulling the dataset off disk into memory, and until that is done it will not take commands. The real fix is on the client side, your app should retry with exponential backoff so a brief LOADING does not take it down. If loading drags on, that is a big persistence file or a slow disk, and an AOF rewrite to shrink the file usually helps the next restart.”
INFO persistence and the loading field.BGREWRITEAOF
LOADING is normal warm-up after a restart while Redis reads its RDB or AOF into memory. Wait for it to finish, and make your client retry transient errors with exponential backoff so it is not fatal. If loading is slow, shrink the AOF with BGREWRITEAOF and check disk speed.
If Redis stays in LOADING far longer than the dataset size suggests, check the logs for AOF or RDB load errors and confirm the disk is not saturated; a corrupt AOF can be repaired with redis-check-aof. You can also submit your error to us for a tailored fix.