Key Concepts
Connection
A configured Redis server endpoint. Supports standalone, Sentinel, and cluster deployments.
Database
Redis organizes keys into numbered databases (0-15 by default). Each database is isolated.
Client
An active connection to Redis. Includes your application connections and admin tools.
Memory
Redis stores all data in memory. Monitor usage to prevent out-of-memory conditions.
Required Permissions
| Action | Permission |
|---|---|
| View overview dashboard | iam:project:infrastructure:redis:read |
| View connected clients | iam:project:infrastructure:redis:read |
Connection Types
| Type | Description |
|---|---|
| Standalone | Single Redis server. Simple setup for development or small deployments. |
| Sentinel | High availability setup with automatic failover. Master-replica topology. |
| Cluster | Distributed Redis across multiple nodes with automatic sharding. |
How to Select a Connection
Your last selected connection is remembered. The page will automatically reconnect on your next visit.
How to Refresh Dashboard Data
Understanding Dashboard Metrics
Server Information
| Metric | Description |
|---|---|
| Version | Redis server version (e.g., 7.2.0) |
| Uptime | Time since Redis server started |
| Total Keys | Number of keys across all databases |
| Clients | Currently connected clients |
| Ops/sec | Instantaneous operations per second |
| Memory | Current memory usage |
Memory Usage
| Metric | Description |
|---|---|
| Used Memory | Current memory allocated by Redis |
| Max Memory | Configured memory limit (0 = unlimited) |
| Peak Memory | Highest memory usage since server start |
| Fragmentation Ratio | Memory fragmentation level |
| Expiring Keys | Keys with TTL set |
Fragmentation Ratio
| Ratio | Meaning | Action |
|---|---|---|
| < 1.0 | Memory underutilized | Normal for small datasets |
| 1.0 - 1.5 | Healthy fragmentation | No action needed |
| > 1.5 | High fragmentation | Consider restarting Redis |
| > 2.0 | Significant fragmentation | Investigate memory issues |
Performance Statistics
| Metric | Description |
|---|---|
| Total Commands | Commands processed since server start |
| Total Connections | Connections received since start |
| Rejected Connections | Connections refused (max clients reached) |
| Blocked Clients | Clients waiting on blocking commands |
Blocked clients are normal when using commands like
BLPOP, BRPOP, or XREAD. Persistent high blocked counts may indicate slow consumers.Understanding Key Distribution
The dashboard shows key counts by data type:| Type | Description | Use Case |
|---|---|---|
| String | Simple key-value pairs | Caching, counters, session data |
| Hash | Field-value maps | Objects, user profiles |
| List | Ordered collections | Queues, recent items |
| Set | Unique unordered members | Tags, unique visitors |
| Sorted Set | Scored members | Leaderboards, priority queues |
| Stream | Append-only log | Event sourcing, messaging |
Understanding Replication
| Field | Description |
|---|---|
| Role | Master (primary) or Slave (replica) |
| Connected Slaves | Number of replicas connected to this master |
| AOF Enabled | Append-Only File persistence status |
| RDB Enabled | Point-in-time snapshot persistence status |
| Mode | Deployment mode (standalone, sentinel, cluster) |
Persistence Types
| Type | Description |
|---|---|
| RDB | Periodic snapshots. Fast recovery, potential data loss between snapshots. |
| AOF | Log every write operation. Slower but more durable. |
| Both | RDB for fast recovery, AOF for durability. Recommended for production. |
How to View Connected Clients
Client Flags
| Flag | Meaning |
|---|---|
| N | Normal client |
| M | Master in MONITOR mode |
| S | Slave connection |
| O | Client in MULTI/EXEC context |
| x | Client in MULTI/EXEC (aborted) |
| b | Waiting on blocking command |
| t | Watched keys modified |
| R | Client in readonly mode |
| P | Pub/Sub subscriber |
Related Pages
| Page | Description |
|---|---|
| Key Browser | Browse, search, and manage Redis keys |
| Slow Log | Analyze slow commands for performance tuning |
| Configuration | View and modify Redis configuration |
| CLI | Execute Redis commands directly |
Troubleshooting
No connection available
No connection available
- No active Redis connections are configured
- Navigate to Settings > Integrations to add a Redis connection
- Verify connection settings (host, port, authentication)
Connection failed
Connection failed
- Verify Redis server is running
- Check network connectivity to the Redis host
- Confirm authentication credentials are correct
- Ensure TLS settings match server configuration
High memory usage
High memory usage
- Check for keys without TTL that should expire
- Review max memory policy (eviction settings)
- Consider enabling memory analysis with
MEMORY DOCTOR - Check for memory fragmentation
High fragmentation ratio
High fragmentation ratio
- Restart Redis to defragment memory
- Enable active defragmentation in Redis 4+
- Check for patterns causing fragmentation (large key deletions)
- Consider memory allocator tuning
Many rejected connections
Many rejected connections
- Check
maxclientsconfiguration - Review if clients are closing connections properly
- Ensure connection pooling is used in applications
- Check for connection leaks
Dashboard not loading
Dashboard not loading
- Verify the selected connection is active
- Check if the Redis server allows INFO command
- Ensure your user has read permission
- Try refreshing the page
FAQ
What's the difference between used and max memory?
What's the difference between used and max memory?
Used Memory is what Redis has allocated for data. Max Memory is the configured limit (0 means no limit). When used memory reaches max memory, Redis applies the eviction policy to free space.
Why is peak memory higher than current memory?
Why is peak memory higher than current memory?
Peak memory shows the highest memory usage since server start. Memory decreases when keys expire or are deleted. Large temporary operations can spike peak memory.
What does ops/sec measure?
What does ops/sec measure?
Operations per second measures the instantaneous rate of Redis commands being processed. This includes all operations: reads, writes, and admin commands.
How are databases different from each other?
How are databases different from each other?
Each database (0-15) is a separate keyspace. Keys in database 0 are independent from keys in database 1. Databases share the same Redis instance and memory.
What causes blocked clients?
What causes blocked clients?
Clients become blocked when using blocking commands like
BLPOP, BRPOP, XREAD, or WAIT. They remain blocked until data becomes available or timeout expires.Why can't I see all connected clients?
Why can't I see all connected clients?
The client list is paginated for performance. Use search to find specific clients by address or name. Very large deployments may have thousands of connections.
What is AOF vs RDB persistence?
What is AOF vs RDB persistence?
RDB creates point-in-time snapshots at intervals. AOF logs every write operation. RDB is faster for recovery; AOF provides better durability. Production systems often use both.
How do I know if replication is healthy?
How do I know if replication is healthy?
For a master, check that expected replicas appear in “Connected Slaves”. For a replica, the role should show “slave” with link status “up”. Replication lag should be minimal.
Best Practices
Regular Monitoring
- Check memory usage trends before reaching limits
- Monitor rejected connections to detect capacity issues
- Track ops/sec during peak hours for capacity planning
- Review fragmentation ratio periodically
Memory Management
- Set appropriate
maxmemorylimits in production - Configure eviction policies based on your use case
- Use TTL on cache keys to prevent unbounded growth
- Monitor expiring keys count
Connection Management
- Use connection pooling in applications
- Set appropriate connection timeouts
- Close idle connections to free resources
- Monitor client count during deployments
Performance Tuning
- Review slow log for optimization opportunities
- Check key distribution for hot spots
- Consider clustering for high-throughput workloads
- Monitor blocked clients for queue processing issues