nodetool tablestats¶

Displays detailed statistics for tables including read/write latencies, SSTable counts, partition sizes, and space usage.

Synopsis¶

nodetool [connection_options] tablestats [options] [--] [keyspace[.table] ...]

Description¶

nodetool tablestats provides comprehensive metrics for tables including:

Read and write operation counts and latencies
SSTable counts and sizes
Partition statistics (count, size)
Bloom filter and compression metrics
Memtable information
Percent repaired (for incremental repair)

This is one of the most important commands for capacity planning and performance analysis.

Arguments¶

Argument	Description
`keyspace`	Display stats for all tables in keyspace
`keyspace.table`	Display stats for specific table

Options¶

Option	Description
`-H, --human-readable`	Display sizes in human-readable format
`-F, --format`	Output format (json, yaml)
`-s, --sort`	Sort by column (read, write, space, etc.)
`-t, --top`	Show only top N tables
`-i, --ignore`	Ignore specific keyspaces/tables

Output Example¶

Keyspace : my_keyspace
        Read Count: 12345678
        Read Latency: 0.234 ms
        Write Count: 9876543
        Write Latency: 0.123 ms
        Pending Flushes: 0

                Table: users
                SSTable count: 12
                Old SSTable count: 0
                Max SSTable size: 128 MB
                Space used (live): 1.5 GB
                Space used (total): 1.8 GB
                Space used by snapshots (total): 256 MB
                Memtable cell count: 12345
                Memtable data size: 64 MB
                Memtable switch count: 234
                Speculative retries: 0
                Local read count: 5678901
                Local read latency: 0.456 ms
                Local write count: 3456789
                Local write latency: 0.123 ms
                Pending flushes: 0
                Percent repaired: 98.5
                Bloom filter false positives: 123
                Bloom filter false ratio: 0.00001
                Bloom filter space used: 24 MB
                Compression ratio: 0.35
                Compacted partition minimum bytes: 64
                Compacted partition maximum bytes: 65536
                Compacted partition mean bytes: 1024
                Average live cells per slice (last five minutes): 12.5
                Maximum live cells per slice (last five minutes): 150
                Average tombstones per slice (last five minutes): 0.5
                Maximum tombstones per slice (last five minutes): 10
                Dropped mutations: 0
                Droppable tombstone ratio: 0.05

Key Metrics Explained¶

Read/Write Latencies¶

Metric	Description	Healthy Range
Local read latency	Average read time	< 5ms (SSD), < 20ms (HDD)
Local write latency	Average write time	< 1ms

High Read Latency

Read latency > 10ms consistently may indicate:

Large partitions
Too many SSTables
Bloom filter issues
Disk I/O problems

SSTable Metrics¶

Metric	Description	Action Threshold
SSTable count	Number of SSTables	> 20-30 warrants investigation
Max SSTable size	Largest SSTable	Varies by compaction strategy
Space used (live)	Data excluding tombstones	Capacity planning
Space used (total)	All data including tombstones	Disk usage

SSTable Count

High SSTable counts increase read latency and memory usage:

STCS: May have many SSTables naturally
LCS: Should be mostly in L1+
TWCS: One per time window

Partition Statistics¶

Metric	Description	Warning Signs
Compacted partition mean bytes	Average partition size	> 100MB is large
Compacted partition maximum bytes	Largest partition	> 100MB needs attention
Live cells per slice	Cells read per query	> 10000 may be problematic

Large Partitions

Large partitions cause:

Heap pressure
GC issues
Uneven load distribution
Repair/compaction problems

Tombstone Metrics¶

Metric	Description
Tombstones per slice	Tombstones scanned per read
Droppable tombstone ratio	Ratio of tombstones eligible for GC

Tombstone Warnings

High tombstones per slice indicates:

Delete-heavy workload
Queries scanning past deletes
Potential for tombstone warnings in logs

Bloom Filter¶

Metric	Description
Bloom filter false positives	Incorrect positive matches
Bloom filter false ratio	False positive rate
Bloom filter space used	Memory for bloom filters

Target false ratio: < 0.01 (1%)

Compression¶

Metric	Description
Compression ratio	Compressed size / uncompressed size

0.3-0.5 is typical for most data
< 0.2 indicates highly compressible data
0.8 may indicate compression isn't helping

Examples¶

All Tables in Keyspace¶

nodetool tablestats my_keyspace

Specific Table¶

nodetool tablestats my_keyspace.users

Human-Readable Sizes¶

nodetool tablestats -H my_keyspace

JSON Output¶

nodetool tablestats -F json my_keyspace.users

Extract Specific Metrics¶

# SSTable counts for all tables
nodetool tablestats | grep -E "Table:|SSTable count"

# Read latencies
nodetool tablestats | grep -E "Table:|Local read latency"

# Space usage
nodetool tablestats | grep -E "Table:|Space used"

Compare Across Nodes¶

for node in node1 node2 node3; do
    echo "=== $node ==="
    ssh "$node" "nodetool tablestats my_keyspace.users | grep -E "SSTable|Space|latency""
done

Common Analysis Patterns¶

Finding Large Partitions¶

nodetool tablestats | grep -A2 "Table:" | grep -E "Table:|maximum bytes"

Tables with max partition > 100MB need data model review.

Identifying Tables Needing Compaction¶

nodetool tablestats | grep -A5 "Table:" | grep -E "Table:|SSTable count"

High SSTable counts may indicate compaction falling behind.

Checking Tombstone Issues¶

nodetool tablestats | grep -E "Table:|tombstones per slice|Droppable"

High tombstones per slice correlates with read latency issues.

Repair Status¶

nodetool tablestats | grep -E "Table:|Percent repaired"

Track incremental repair progress.

Performance Baselines¶

Healthy Metrics¶

Metric	Target
Local read latency	< 5ms
Local write latency	< 1ms
SSTable count	< 20-30
Bloom filter false ratio	< 0.01
Tombstones per slice	< 100
Partition mean bytes	< 10MB

Warning Thresholds¶

Metric	Warning Level
Local read latency	> 20ms
SSTable count	> 50
Bloom filter false ratio	> 0.1
Tombstones per slice	> 1000
Partition maximum bytes	> 100MB

Troubleshooting with tablestats¶

Slow Reads¶

Check these metrics: 1. SSTable count - too many? 2. Partition size - too large? 3. Tombstones per slice - too many? 4. Bloom filter false ratio - ineffective?

High Disk Usage¶

Check: 1. Space used by snapshots - orphaned snapshots? 2. Droppable tombstone ratio - need compaction? 3. Compression ratio - compression working?

Memory Issues¶

Check: 1. Bloom filter space - large tables consume heap 2. Memtable size - appropriate for workload? 3. Partition sizes - large partitions stress heap

Command	Relationship
tpstats	Thread pool statistics
proxyhistograms	Coordinator-level latencies
compactionstats	Active compaction information
info	Node-level overview