An Examination of Monitoring Metrics: Part 3 Redis

Ready to streamline your complex Zabbix operations?

I’ve distilled the most valuable insights from this blog into one essential guide. Take full control of your environment with the Zabbix 7 Enterprise Optimization Handbook [Mastering Hybrid Infrastructure Monitoring with a Kubernetes First Approach].

👉 Get your PDF copy here: https://jikimy.gumroad.com/l/zabbixmaster

🧭 Looking for related posts? Search for “An Examination of Monitoring Metrics” in the search bar.

1. Memory Metrics

used_memory

• The actual memory consumed by the Redis process.
• A sharp increase in usage compared to physical memory may indicate an OOM (Out Of Memory) risk.
• Always verify maxmemory settings and eviction policies.

mem_fragmentation_ratio

• Memory fragmentation ratio.
• Close to 1.0 is normal.
• If ≥ 1.5, fragmentation is severe → consider restart or RDB/AOF rewrite.

evicted_keys

• Number of keys forcibly removed when maxmemory is exceeded.
• An increase implies rising cache miss probability.
• Review eviction policies (noeviction, allkeys-lru, etc.).

2. Performance Metrics

instantaneous_ops_per_sec

• Number of commands processed per second (QPS).
• Useful for identifying traffic spikes.
• Watch for sudden increases or drops against baseline.

slowlog

• Number of slow query entries recorded in Redis per second.
• A value near 0 is normal.
• Persistent growth indicates blocking commands or large dataset operations in applications.

3. Connection Metrics

blocked_clients

• Number of clients waiting due to blocking commands (BRPOP, BLPOP, etc.).
• A surge may indicate queue processing bottlenecks.

connected_clients

• Current number of connected clients.
• Compare with application connection pool configuration.
• Approaching maxclients risks new connection failures.

rejected_connections

• Number of connections rejected due to exceeding concurrency limits.
• Frequent increases suggest adjusting client pool settings.

4. Network Metrics

total_net_input_bytes / total_net_output_bytes

• Cumulative inbound/outbound data volume.
• Useful for identifying network bandwidth trends.
• Spikes at specific times may indicate large value SET/GET operations.

5. Persistence Metrics

(Persistence: the property of data being safely preserved beyond memory to disk)

rdb_last_bgsave_status

• Status of the last RDB snapshot (success/failure).
• Failures may be due to insufficient disk space or permission errors.

aof_last_bgrewrite_status

• Status of the last AOF rewrite (success/failure).
• Monitor disk usage when AOF file size grows rapidly.

rdb_changes_since_last_save

• Number of key changes since the last save.
• If it grows excessively beyond the save interval, data loss risk increases in case of failure.

6. Cache Efficiency Metrics

keyspace_hits / keyspace_misses

• Cache Hit Ratio = hits / (hits + misses) * 100
• Should ideally remain ≥ 90.
• A lower ratio suggests a need to review cache policies or data structures.

⚠ Note: The default Redis template for Zabbix agent2 does not include a Cache Hit Ratio item.
You must create a calculated item based on existing values:

last(//redis.stats.keyspace_hits) /
( last(//redis.stats.keyspace_hits) + last(//redis.stats.keyspace_misses) ) * 100

✅ Operational Guidelines

• Prioritize checking memory usage and fragmentation ratio.
• Monitor connection counts and rejections to identify client connection issues.
• Use ops/sec and slowlog together to detect performance degradation.
• Watch cache hit ratio and evicted_keys growth as signals for reviewing cache policies.

🛠 마지막 수정일: 2025.12.22