Cassandra Metrics Reference

Apache Cassandra exposes 500+ metrics via JMX using the Dropwizard Metrics library. This reference provides comprehensive documentation of all metrics with recommended thresholds.

Metrics Overview

Cassandra metrics follow this naming pattern:

org.apache.cassandra.metrics:type={Type},keyspace={ks},scope={scope},name={MetricName}

Metric Types

Type	Description
Timer	Latency distribution (count, mean, percentiles)
Counter	Cumulative count
Gauge	Point-in-time value
Histogram	Value distribution
Meter	Rate of events

---

Client Request Metrics

Type: ClientRequest

Read Latency

ObjectName: org.apache.cassandra.metrics:type=ClientRequest,scope=Read,name=Latency

Attribute	Type	Description
Count	long	Total reads
Mean	double	Mean latency (μs)
50thPercentile	double	p50 latency
75thPercentile	double	p75 latency
95thPercentile	double	p95 latency
99thPercentile	double	p99 latency
999thPercentile	double	p999 latency
Max	double	Maximum latency
OneMinuteRate	double	Reads per second

Thresholds:

Percentile	Warning	Critical
p50	5ms	20ms
p99	50ms	500ms
p999	200ms	2000ms

Write Latency

ObjectName: org.apache.cassandra.metrics:type=ClientRequest,scope=Write,name=Latency

Attribute	Type	Description
Count	long	Total writes
Mean	double	Mean latency (μs)
50thPercentile	double	p50 latency
99thPercentile	double	p99 latency
OneMinuteRate	double	Writes per second

Thresholds:

Percentile	Warning	Critical
p50	2ms	10ms
p99	10ms	100ms

Timeouts

ObjectName: org.apache.cassandra.metrics:type=ClientRequest,scope={Read|Write},name=Timeouts

Attribute	Description
Count	Total timeout count
OneMinuteRate	Timeouts per second

Alert: Any timeouts warrant investigation.

Unavailables

ObjectName: org.apache.cassandra.metrics:type=ClientRequest,scope={Read|Write},name=Unavailables

Attribute	Description
Count	Total unavailable errors
OneMinuteRate	Unavailables per second

Alert: Indicates insufficient replicas available for requested consistency level.

Failures

ObjectName: org.apache.cassandra.metrics:type=ClientRequest,scope={Read|Write},name=Failures

Attribute	Description
Count	Total failures
OneMinuteRate	Failures per second

---

Table Metrics

Type: Table

Per-keyspace and per-table metrics.

Read/Write Latency (Per Table)

ObjectName: org.apache.cassandra.metrics:type=Table,keyspace={ks},scope={table},name=ReadLatency

Same attributes as ClientRequest latency but scoped to specific table.

Live SSTable Count

ObjectName: org.apache.cassandra.metrics:type=Table,keyspace={ks},scope={table},name=LiveSSTableCount

Attribute	Description
Value	Number of live SSTables

Threshold: > 20-30 SSTables may indicate compaction lag.

SSTable Size

ObjectName: org.apache.cassandra.metrics:type=Table,keyspace={ks},scope={table},name=TotalDiskSpaceUsed

Attribute	Description
Count	Bytes on disk

Memtable Size

ObjectName: org.apache.cassandra.metrics:type=Table,keyspace={ks},scope={table},name=MemtableLiveDataSize

Attribute	Description
Value	Bytes in memtable

Partition Size

ObjectName: org.apache.cassandra.metrics:type=Table,keyspace={ks},scope={table},name=EstimatedPartitionSizeHistogram

Attribute	Description
Value	Histogram array of partition sizes

Alert: Partitions > 100MB are problematic.

Tombstone Scans

ObjectName: org.apache.cassandra.metrics:type=Table,keyspace={ks},scope={table},name=TombstoneScannedHistogram

Attribute	Description
Value	Histogram of tombstones scanned per read

Threshold: > 1000 tombstones/read indicates query optimization needed.

Bloom Filter

ObjectName: org.apache.cassandra.metrics:type=Table,keyspace={ks},scope={table},name=BloomFilterFalseRatio

Attribute	Description
Value	False positive ratio (0.0-1.0)

Threshold: > 0.01 may indicate need for larger bloom filters.

Speculative Retries

ObjectName: org.apache.cassandra.metrics:type=Table,keyspace={ks},scope={table},name=SpeculativeRetries

Attribute	Description
Count	Number of speculative retries

---

Compaction Metrics

Type: Compaction

Pending Tasks

ObjectName: org.apache.cassandra.metrics:type=Compaction,name=PendingTasks

Attribute	Description
Value	Pending compaction tasks

Thresholds:

Level	Value
Normal	< 5
Warning	5-20
Critical	> 50

Completed Tasks

ObjectName: org.apache.cassandra.metrics:type=Compaction,name=CompletedTasks

Attribute	Description
Value	Completed compactions

Bytes Compacted

ObjectName: org.apache.cassandra.metrics:type=Compaction,name=BytesCompacted

Attribute	Description
Count	Total bytes compacted
OneMinuteRate	Compaction throughput

Total Compactions

ObjectName: org.apache.cassandra.metrics:type=Compaction,name=TotalCompactionsCompleted

Attribute	Description
Count	Total completed compactions
OneMinuteRate	Compactions per second

---

Thread Pool Metrics

Type: ThreadPools

Stage Pools

Pool	Path	Description
ReadStage	request	Processes reads
MutationStage	request	Processes writes
CounterMutationStage	request	Counter writes
ViewMutationStage	request	MV updates
GossipStage	internal	Gossip processing
CompactionExecutor	internal	Compaction
MemtableFlushWriter	internal	Memtable flush
MemtablePostFlush	internal	Post-flush
Native-Transport-Requests	transport	CQL requests

Active Tasks

ObjectName: org.apache.cassandra.metrics:type=ThreadPools,path={path},scope={stage},name=ActiveTasks

Attribute	Description
Value	Currently active tasks

Pending Tasks

ObjectName: org.apache.cassandra.metrics:type=ThreadPools,path={path},scope={stage},name=PendingTasks

Attribute	Description
Value	Tasks waiting in queue

Alert: Pending tasks > 0 for extended periods indicates overload.

Blocked Tasks

ObjectName: org.apache.cassandra.metrics:type=ThreadPools,path={path},scope={stage},name=CurrentlyBlockedTasks

Attribute	Description
Count	Total blocked tasks
Value	Currently blocked

Alert: Any blocked tasks indicate severe overload.

Completed Tasks

ObjectName: org.apache.cassandra.metrics:type=ThreadPools,path={path},scope={stage},name=CompletedTasks

Attribute	Description
Value	Total completed tasks

---

Cache Metrics

Type: Cache

Key Cache

Metric	ObjectName Suffix	Description
Size	scope=KeyCache,name=Size	Entries in cache
Capacity	scope=KeyCache,name=Capacity	Max entries
Hits	scope=KeyCache,name=Hits	Cache hits
Requests	scope=KeyCache,name=Requests	Total requests
HitRate	scope=KeyCache,name=HitRate	Hit ratio

Target: > 85% hit rate.

Row Cache

Metric	ObjectName Suffix	Description
Size	scope=RowCache,name=Size	Entries in cache
Capacity	scope=RowCache,name=Capacity	Max entries
Hits	scope=RowCache,name=Hits	Cache hits
HitRate	scope=RowCache,name=HitRate	Hit ratio

Counter Cache

Metric	ObjectName Suffix	Description
Size	scope=CounterCache,name=Size	Entries in cache
HitRate	scope=CounterCache,name=HitRate	Hit ratio

Chunk Cache (Off-Heap)

ObjectName: org.apache.cassandra.metrics:type=Cache,scope=ChunkCache,name={metric}

Metric	Description
Size	Cached chunks
Misses	Cache misses
MissLatency	Miss latency

---

Dropped Messages Metrics

Type: DroppedMessage

ObjectName: org.apache.cassandra.metrics:type=DroppedMessage,scope={type},name=Dropped

Type	Description
MUTATION	Dropped writes
READ	Dropped reads
READ_REPAIR	Dropped read repairs
RANGE_SLICE	Dropped range queries
COUNTER_MUTATION	Dropped counter writes
HINT	Dropped hints
PAGED_RANGE	Dropped paged queries

Attribute	Description
Count	Total dropped
OneMinuteRate	Drops per second

Alert: Any dropped messages indicate severe issues.

---

Storage Metrics

Type: Storage

Disk Space

ObjectName: org.apache.cassandra.metrics:type=Storage,name=Load

Attribute	Description
Count	Total bytes on disk

Exceptions

ObjectName: org.apache.cassandra.metrics:type=Storage,name=Exceptions

Attribute	Description
Count	Storage exceptions

---

Connection Metrics

Type: Connection

Client Connections

ObjectName: org.apache.cassandra.metrics:type=Client,name=connectedNativeClients

Attribute	Description
Value	Connected CQL clients

Inter-node Connections

ObjectName: org.apache.cassandra.metrics:type=Connection,scope={host},name=*

Metric	Description
CommandPendingTasks	Pending requests
CommandCompletedTasks	Completed requests
CommandDroppedTasks	Dropped requests

---

Streaming Metrics

Type: Streaming

ObjectName: org.apache.cassandra.metrics:type=Streaming,scope={host},name=*

Metric	Description
IncomingBytes	Bytes received
OutgoingBytes	Bytes sent

---

JVM Metrics

Standard JVM MBeans:

Heap Memory

ObjectName: java.lang:type=Memory

Attribute	Description
HeapMemoryUsage.used	Current heap usage
HeapMemoryUsage.max	Maximum heap

Threshold: > 70% warning, > 85% critical.

Garbage Collection

ObjectName: java.lang:type=GarbageCollector,name={collector}

Attribute	Description
CollectionCount	Total collections
CollectionTime	Total collection time (ms)

---

Metrics Categories Summary

Category	Key Metrics	Alert Focus
Client Requests	Latency, Timeouts, Unavailables	Response times
Tables	SSTable count, Partition size	Data model
Compaction	Pending tasks	Maintenance lag
Thread Pools	Pending, Blocked	Capacity
Caches	Hit rate	Efficiency
Dropped Messages	All types	Overload

---

Next Steps

• MBeans Reference - MBean operations
• Monitoring Guide - End-to-end monitoring
• Alerting - Alert configuration
• Troubleshooting - Using metrics for diagnosis