Redis running out of memory

When Redis nears its memory limit, flushing it or lifting the limit clears the alert, but a flush deletes data an application may still need, and without a limit nothing keeps memory in check. 2501 finds what is filling the memory, deletes no data, and leaves the capacity decision to a person.

Depending on how it is set up, a full Redis either refuses writes or quietly drops keys to make room, and the application behind it starts to fail, sometimes with no obvious error.

What 2501 does

The NoSQL specialty takes the alert from your ticketing system and establishes how Redis is set up, on its own, with automatic failover or as a cluster, since each fails differently. It pins the pressure to its cause: memory that jumped or grew steadily, a few very large keys, or keys that should have expired and never did. It also checks that replicas and backups to disk are healthy.

It reports the cause with the evidence and a recommended next step. Capacity problems are escalated for planning, never solved by deleting data.

You stay in control

• You choose the level: investigate only, act after a person confirms, or act on its own within your rules. The level is set per specialty, not per ticket.
• No data is deleted to make room: it never flushes a database, bulk-deletes keys or removes the memory limit.
• Capacity is your call: more memory is a planning decision, so it goes to a person with the growth and its cause.
• Failover and backups stay yours: no failover or cluster change without approval, and backups to disk are never switched off.

Frequently asked

Why not just flush the cache? A flush deletes everything in that database, including data an application may still need, so 2501 never flushes Redis.

What if keys that should expire never do? Memory keeps growing, and a bigger limit only delays the next alert. 2501 reports those keys as the cause, so they can be fixed where they are written.

Related

• NoSQL Manager
• Read-only AI investigation