nodetool getcompactionthroughput¶

Displays the current compaction throughput limit.

Synopsis¶

nodetool [connection_options] getcompactionthroughput

Description¶

nodetool getcompactionthroughput returns the maximum rate at which compaction operations can write data, measured in megabytes per second (MB/s). This throttle helps prevent compaction from consuming excessive disk I/O and impacting read/write performance.

Compaction throughput affects:

SSTable merge operations - How fast SSTables are combined
Disk I/O utilization - Balance between compaction and client traffic
Compaction backlog - How quickly pending compactions complete
Read performance - Fewer SSTables means faster reads

Output¶

Standard Output¶

Current compaction throughput: 64 MB/s

Output Values¶

Value	Meaning
`64 MB/s`	Default throughput limit
`0 MB/s`	Unlimited (no throttling)
`16-256 MB/s`	Typical operational range

Examples¶

Basic Usage¶

nodetool getcompactionthroughput

Output:

Current compaction throughput: 64 MB/s

Check Remote Node¶

ssh 192.168.1.100 "nodetool getcompactionthroughput"

Cluster-Wide Check¶

for node in node1 node2 node3; do
    echo -n "$node: "
    ssh "$node" "nodetool getcompactionthroughput"
done

Extract Value Only¶

nodetool getcompactionthroughput | awk '{print $4}'

Output:

Understanding Compaction Throughput¶

How Throttling Works¶

Cassandra limits the rate at which compaction writes output SSTables:

Compaction reads input SSTables
Merges and processes data in memory
Writes output SSTables at throttled rate
Rate is controlled by compaction throughput setting

Impact of Different Values¶

Throughput	Compaction Speed	I/O Impact	Use Case
16 MB/s	Very slow	Minimal	Heavy read load
32 MB/s	Slow	Low	Normal production
64 MB/s	Moderate	Moderate	Default/balanced
128 MB/s	Fast	High	Maintenance windows
256+ MB/s	Very fast	Very high	Dedicated compaction
0 (unlimited)	Maximum	Maximum	Recovery/emergency

When to Use¶

Routine Monitoring¶

Check current throughput setting:

nodetool getcompactionthroughput

Before Adjusting¶

Verify current value before making changes:

# Check current
nodetool getcompactionthroughput

# Adjust if needed
nodetool setcompactionthroughput 128

# Verify change
nodetool getcompactionthroughput

During Compaction Issues¶

When investigating compaction performance:

# Check throughput limit
nodetool getcompactionthroughput

# Check current compaction activity
nodetool compactionstats

# Check if throughput is the bottleneck
# (compare active rate vs configured limit)

Cluster Configuration Audit¶

Verify consistency across nodes:

#!/bin/bash
# Check compaction throughput on all nodes

echo "=== Compaction Throughput Audit ==="
for node in $(nodetool status | grep "^UN" | awk '{print $2}'); do
    throughput=$(ssh "$node" "nodetool getcompactionthroughput 2>/dev/null | awk '{print $4}')"
    echo "$node: $throughput MB/s"
done

Correlation with Compaction Stats¶

Monitoring Actual Throughput¶

# Check configured limit
nodetool getcompactionthroughput

# Check actual compaction activity
nodetool compactionstats

Example compactionstats output:

pending tasks: 5
- system_schema.tables: 0
- my_keyspace.my_table: 5

Active compaction:
   Compacting (my_keyspace.my_table)
   Progress: 45.2% (2.3/5.1 GB)
   Throughput: 62.5 MB/s

If "Throughput" consistently matches the configured limit, compaction is being throttled.

Signs Throughput is Too Low¶

Indicator	Symptom
Growing pending tasks	`nodetool compactionstats` shows increasing backlog
High SSTable count	`nodetool tablestats` shows many SSTables
Read latency increase	More SSTables to scan for reads
Disk usage growth	Uncompacted data accumulating

Signs Throughput is Too High¶

Indicator	Symptom
I/O saturation	High disk utilization (iostat)
Read/write latency	Client operations slowed
Dropped messages	`nodetool tpstats` shows drops
CPU pressure	Compaction using too much CPU

Default Values¶

By Version¶

Cassandra Version	Default	Notes
4.x	64 MB/s	Recommended default
3.x	16 MB/s	Conservative default
2.x	16 MB/s	Legacy default

Configuration File¶

In cassandra.yaml:

# Throttle for compaction in MB/s. Set to 0 to disable throttling.
compaction_throughput_mb_per_sec: 64

Setting	Purpose	Command
Compaction throughput	SSTable write rate limit	`getcompactionthroughput` / `setcompactionthroughput`
Stream throughput	Node-to-node transfer rate	`getstreamthroughput` / `setstreamthroughput`
Concurrent compactors	Parallel compaction threads	`getconcurrentcompactors` / `setconcurrentcompactors`

Understanding the Relationship¶

# Check all compaction-related settings
echo "=== Compaction Settings ==="
echo "Throughput: $(nodetool getcompactionthroughput)"
echo "Concurrent compactors: $(nodetool getconcurrentcompactors)"
echo ""
echo "=== Current Activity ==="
nodetool compactionstats

Effective throughput = throughput_per_compactor × concurrent_compactors

Tuning Guidelines¶

When to Increase¶

Compaction backlog growing
Many pending compaction tasks
During maintenance windows
SSTable count too high

# Check current
nodetool getcompactionthroughput

# Increase for maintenance
nodetool setcompactionthroughput 128

# Monitor progress
watch nodetool compactionstats

When to Decrease¶

High client latencies during compaction
I/O contention observed
Disk utilization too high
During peak traffic periods

# Check current
nodetool getcompactionthroughput

# Reduce to minimize impact
nodetool setcompactionthroughput 32

# Verify client latencies improve
nodetool proxyhistograms

Monitoring Script¶

#!/bin/bash
# compaction_monitor.sh - Monitor compaction throughput and activity

while true; do
    clear
    echo "=== Compaction Monitor $(date) ==="
    echo ""

    # Configured limit
    echo "Configured throughput: $(nodetool getcompactionthroughput)"
    echo ""

    # Current activity
    echo "Current compaction activity:"
    nodetool compactionstats

    echo ""
    echo "Press Ctrl+C to exit"
    sleep 10
done

Troubleshooting¶

Throughput Shows 0¶

nodetool getcompactionthroughput
# Output: Current compaction throughput: 0 MB/s

This means compaction is unlimited (no throttling). This is intentional in some configurations but may cause I/O saturation.

# Set a reasonable limit
nodetool setcompactionthroughput 64

Value Doesn't Persist After Restart¶

Runtime changes with setcompactionthroughput don't persist. Update cassandra.yaml:

compaction_throughput_mb_per_sec: 64

Different Values Across Nodes¶

# Audit cluster
for node in node1 node2 node3; do
    echo "$node: $(ssh "$node" "nodetool getcompactionthroughput | awk '{print $4}')""
done

# Output:
# node1: 64
# node2: 32  # Different!
# node3: 64

# Standardize
ssh node2 "nodetool setcompactionthroughput 64"

Best Practices¶

Compaction Throughput Guidelines

Monitor regularly - Include in operational dashboards
Consistent across cluster - Use same setting on all nodes
Adjust for workload - Lower during peak, higher during maintenance
Balance with streaming - Consider total I/O budget
Document changes - Log when and why throughput was adjusted
Persist settings - Update cassandra.yaml for permanent changes

Command	Relationship
setcompactionthroughput	Modify throughput limit
compactionstats	View active compactions
compactionhistory	View past compactions
setstreamthroughput	Set streaming rate
tablestats	SSTable counts per table