Redis Integration

Monitor Redis memory pressure, command performance, keyspace behaviour, persistence, slow commands, replication and cluster health from a single Watchlog agent.

The agent runs redis-cli once per collection cycle and derives everything from that single round-trip.

Supported Platforms

• Ubuntu
• Windows
• Source Code
• Docker Container
• Kubernetes

Requirements

redis-cli --version

Authentication

With ACLs, a read-only monitoring user is enough:

ACL SETUSER watchlog_monitor on >your_password ~* &* +@read +info +config|get +slowlog|get +client|list +cluster|info +cluster|nodes +replicaof|no

Or use the built-in read-only pattern if your policy prefers it. The commands Watchlog actually issues are INFO, INFO commandstats, SLOWLOG GET, CONFIG GET slowlog-*, CLUSTER INFO and CLUSTER NODES — nothing else.

What Watchlog Collects

Server and memory Version, mode, role, uptime, connected clients and blocked clients, used memory, peak memory, maxmemory and the eviction policy, RSS, memory fragmentation ratio, and Lua/script memory.

Commands (throttled — see intervals) Per command: calls, total and per-call microseconds, and rejected/failed counts where the server reports them. Ranked by total time consumed, so a cheap command called constantly ranks above a rare expensive one.

Keyspace Per database: keys, keys with a TTL, and average TTL. Plus cluster-wide keyspace hits and misses with the derived interval hit rate — not the since-boot ratio, which on a server up for months can no longer move.

Persistence RDB: last save time, changes since last save, last background save status and duration. AOF: enabled state, rewrite in progress, last rewrite status, current size and base size.

Slow commands From SLOWLOG GET: id, timestamp, duration, the command with arguments redacted, and the client address and name. The agent also reads CONFIG GET slowlog-* so the dashboard can tell you whether your configured threshold makes sense — it never writes that configuration.

Replication Role, connected replicas with their offsets and lag, master link status on a replica, and replication backlog state.

Cluster (when cluster mode is on) Cluster state, slots assigned / OK / PFAIL / FAIL, known nodes, cluster size, and per-node role and link state.

Health Score and Insights

A 0–100 score computed server-side, with a written reason on every deduction:

Insights add the interpretation: an eviction spike traced to a policy, a hit rate that dropped after a deploy, a replica that is connected but not streaming, and slot coverage that would make part of the keyspace unreachable.

Slow Commands

SLOWLOG is on by default in Redis, unlike MongoDB's profiler or Elasticsearch's slow logs — so this generally works with no server-side setup.

Watchlog reads it but never resets it. SLOWLOG RESET is destructive to anyone else reading the same log, so the agent tracks a high-water mark per instance instead: the same entry is never shipped twice, and never re-shipped after a restart.

Argument redaction

Command arguments are redacted on your host, before they leave it, and the rule is allow-list shaped rather than deny-list shaped — an argument is redacted unless it is provably safe to keep:

Everything else becomes [REDACTED]. A SET session:abc <token> stores the key and the command, never the value. RESTORE never ships its serialized payload.

Collection Intervals

Safety

Watchlog is read-only. It never runs any of the following:

KEYS * · MONITOR · DEBUG SLEEP · FLUSHALL · FLUSHDB · CONFIG SET · SLOWLOG RESET · SCRIPT FLUSH · any write command

KEYS * and MONITOR deserve special mention: both are O(N) blocking operations that have taken down production Redis instances used exactly this way by other monitoring tools. Watchlog reads key counts from INFO keyspace, which is O(1).

Advanced Options

{
  "service": "redis",
  "monitor": true,
  "host": "127.0.0.1",
  "port": "6379",
  "username": "",
  "password": "your_password",
  "db": "",
  "tls": false,

  "advanced": {
    "enabled": true,
    "commands": true,
    "keyspace": true,
    "replication": true,
    "cluster": true,
    "maxCommands": 200,
    "commandsIntervalSeconds": 60
  },

  "slowlog": {
    "enabled": true,
    "limit": 128,
    "maxPerScrape": 100
  }
}

Multiple Instances

Add one entry per instance. Each is identified by its host:port:

[
  {
    "service": "redis",
    "monitor": true,
    "host": "127.0.0.1",
    "port": "6379",
    "password": "your_password"
  },
  {
    "service": "redis",
    "monitor": true,
    "host": "127.0.0.1",
    "port": "6380",
    "password": "your_password"
  }
]

Troubleshooting