nodetool flush¶

Flushes memtables from memory to SSTables on disk for one or more tables.

Synopsis¶

nodetool [connection_options] flush [--] [keyspace [table ...]]

Description¶

nodetool flush forces an immediate flush of memtable data to disk as SSTables. Memtables hold recent writes in memory; flushing writes this data to immutable SSTable files.

Arguments¶

Argument	Description
`keyspace`	Keyspace to flush. If omitted, flushes all keyspaces
`table`	Specific table(s) to flush. If omitted, flushes all tables in keyspace

Examples¶

Flush All Keyspaces¶

nodetool flush

Flushes all memtables for all keyspaces on the node.

Flush Specific Keyspace¶

nodetool flush my_keyspace

Flushes all tables in my_keyspace.

Flush Specific Table¶

nodetool flush my_keyspace my_table

Flushes only my_table in my_keyspace.

Flush Multiple Tables¶

nodetool flush my_keyspace table1 table2 table3

Flushes multiple specific tables.

When to Use¶

Before Taking Snapshots¶

Required Before Backups

Always flush before creating snapshots to ensure all data is on disk:

nodetool flush my_keyspace
nodetool snapshot -t backup_$(date +%Y%m%d) my_keyspace

Without flushing, recent writes in memtables will not be included in the snapshot.

Before Stopping Cassandra¶

nodetool flush
nodetool drain
sudo systemctl stop cassandra

Drain Includes Flush

nodetool drain automatically flushes all memtables. Explicit flush before drain is optional but makes the drain faster.

Before Major Compaction¶

nodetool flush my_keyspace
nodetool compact my_keyspace

Ensures all data is in SSTables before compaction.

Before Upgrading SSTables¶

nodetool flush
nodetool upgradesstables

Ensures no data remains in memtables before SSTable upgrade.

When NOT to Use¶

Unnecessary Flushing

Avoid excessive flushing:

Frequent flushes: Creates many small SSTables, increasing compaction overhead
During normal operations: Let Cassandra manage memtable flushes automatically
Under heavy write load: Adds disk I/O pressure

Cassandra's Automatic Flushing¶

Cassandra automatically flushes memtables when:

Condition	Configuration
Memtable size threshold	`memtable_heap_space`
Commit log segment full	`commitlog_segment_size`
Time-based	`memtable_flush_period_in_ms` (if configured)
Memory pressure	When approaching heap limits

Impact on Operations¶

Resource Usage¶

Resource	Impact
Disk I/O	Write burst during flush
Memory	Temporary increase while flushing
CPU	Compression during SSTable write

During Flush¶

Write operations continue to new memtable
Read operations may read from both memtable and SSTables
Flush is generally quick for reasonably-sized memtables

SSTable Creation¶

Each flush creates new SSTable files:

/var/lib/cassandra/data/my_keyspace/my_table-<uuid>/
├── nb-1-big-Data.db
├── nb-1-big-Index.db
├── nb-1-big-Filter.db
└── ...

Compaction Follows

New SSTables from flushes eventually get compacted according to the table's compaction strategy.

Monitoring Flush Operations¶

Check Pending Flushes¶

nodetool tpstats | grep -i flush

Shows flush-related thread pool activity.

Check Memtable Status¶

nodetool tablestats my_keyspace.my_table | grep -i memtable

Shows current memtable size and flush statistics.

Flush vs. Drain¶

Operation	Scope	Additional Actions
`flush`	Memtables only	None
`drain`	Memtables + connections	Disables gossip and native transport, stops accepting writes

Use drain for graceful shutdown; use flush for data persistence while keeping node operational.

Common Issues¶

Flush Takes Too Long¶

Large memtables increase flush time:

Cause	Solution
Large memtable size	Reduce `memtable_heap_space`
Slow disk I/O	Improve storage performance
Heavy write load	Consider flush during lower traffic

Out of Disk Space During Flush¶

Disk Space Required

Ensure sufficient disk space before flushing:

Flush creates new SSTable files
Space needed: approximately memtable size × compression ratio
Monitor disk usage: df -h /var/lib/cassandra

Many Small SSTables After Flush¶

Frequent flushes create fragmented SSTables:

# Check SSTable count
nodetool tablestats my_keyspace.my_table | grep "SSTable count"

Let compaction consolidate files, or investigate why flushes are frequent.

Best Practices¶

Flush Guidelines

Before backups: Always flush before snapshots
Before shutdown: Use drain which includes flush
One node at a time: For cluster-wide operations
Monitor disk space: Ensure capacity for new SSTables
Avoid in scripts loops: Don't repeatedly flush same tables

Command	Relationship
drain	Flush + disable node
snapshot	Create backup after flush
compact	Consolidate SSTables
tablestats	Check memtable and SSTable stats
tpstats	Monitor flush thread pool