Writes started failing with OOM
Used memory crossed the maxmemory line and the eviction policy either cannot free space or is set to noeviction. Raising the limit without picking a policy only delays the next outage.
Sized correctly, eviction tuned
Working with Redis?
Redis is quick until it isn't. One day it is serving millions of requests a second, the next it is refusing writes because maxmemory filled, or a replica went read-only, or nobody is sure whether to stay on Redis or move to Valkey after the licence change. We work out what is actually happening and fix that.
What it is
Redis, short for Remote Dictionary Server, was created by Salvatore Sanfilippo in 2009. It began as an in-memory key-value store and grew into a richly-typed data store and messaging system used for caches, session stores, rate limiting, queues and real-time counters.
Redis was BSD-licensed until March 2024, when Redis Ltd. moved the source to RSALv2 and SSPL, which are no longer OSI-approved open source. The Linux Foundation then forked Redis 7.2.4 as Valkey, now backed by AWS, Google Cloud, Oracle and Ericsson. Redis Open Source is still available under the new licences for most self-hosting use cases.
The honest version
No technology is all upside. This is where Redis earns its place and where it tends to cause trouble.
Sound familiar?
The Redis problems that bring people to us, and what you have once each one is fixed.
Used memory crossed the maxmemory line and the eviction policy either cannot free space or is set to noeviction. Raising the limit without picking a policy only delays the next outage.
Sized correctly, eviction tuned
The March 2024 licence change made Redis non-OSI open source. Valkey is the Linux Foundation fork and is now the default in the big clouds, but drop-in compatibility has limits.
A documented decision either way
RDB save failing on a full disk or a read-only mount, or AOF rewrite stuck. Redis refuses writes until the operator either fixes the disk or disables stop-writes-on-bgsave-error.
Writes flowing, disk watched
After a failover the roles swapped and the client kept talking to the same address. Sentinel or a proxy is meant to prevent this, but only if clients honour it.
Clients follow the primary
The client library does not follow MOVED and ASK redirects, or code uses multi-key commands across slots. Works on single-node Redis, breaks under Cluster.
Cluster-aware client, hash tags where needed
Redis is single-threaded for commands. A SMEMBERS on a million-member set, a KEYS on production or a big Lua script stalls every other client for the duration.
Slow commands replaced, cursor scans in place
Over time, data that only lives in Redis quietly becomes the only copy. A node reboot wipes it, and nobody has a current RDB or AOF to restore from.
Backups in place, source of truth clear
Nobody has the credentials, nothing is monitored, and the config was last edited years ago. We take it over, document it and put alerts on the parts that actually matter.
Owned, monitored, documented
What we do
Call us for one OOM, or hand us the whole fleet.
How it works
Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.
We fix the cause, not the symptom. If something is down, getting it back up comes first.
Backups, monitoring, updates and access, so the same problem can't quietly come back.
You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.
Often found together
We work on the whole stack, so the problem can be wherever it actually is.
It depends on how you use Redis and who hosts it. For teams on AWS ElastiCache, Google Memorystore or Azure Cache, the managed services are moving to Valkey anyway, so you inherit it. For self-hosters who need true open source, Valkey is now the BSD-licensed successor and is a reasonable default. Teams using Redis Stack modules or Redis Enterprise usually stay on Redis.
The used memory has crossed the maxmemory setting and the eviction policy either cannot free enough space or is set to noeviction. The fix is rarely just a bigger server. We work out which keys are using the memory, whether TTLs are missing, and which eviction policy matches how the application actually reads and writes.
Yes. We design the topology for your workload, configure quorum and failover timeouts that make sense, test failover before it is needed, and make sure the client libraries actually follow MOVED and ASK redirects or the Sentinel primary address.
Usually a mix of keys written without a TTL, abandoned queues from old features, and session data from users who never came back. We audit what is in the keyspace, add TTLs where they belong, and set an eviction policy so memory does not silently fill again.
Yes. We pick up undocumented Redis fleets, work out what each instance is used for, document the config and failure modes, and put monitoring in place for the things that actually cause outages.
Estimates are free. Send over INFO output and the slowlog, and we quote a plan before anyone touches the server.
Free estimate
Tell us what you're running and what's going wrong. A person reads it and comes back with an estimate.
Your message is with a real person, and we'll get back to you soon.
Rather talk now? Call us: