apache/cassandra · critical · IOException

Insufficient disk space to store %s

Error message

Insufficient disk space to store %s

What it means

RangeAwareSSTableWriter's constructor consults Directories.DataDirectory positions to decide where to place the new SSTable; when no data directory has enough free space for the estimated totalSize, getWriteableLocation returns null and the constructor throws IOException 'Insufficient disk space to store <pretty-printed size>'. This is the disk-space guard on the write path before any file is created.

Source

Thrown at src/java/org/apache/cassandra/io/sstable/RangeAwareSSTableWriter.java:74

    public RangeAwareSSTableWriter(ColumnFamilyStore cfs, long estimatedKeys, long repairedAt, TimeUUID pendingRepair, boolean isTransient, SSTableFormat<?, ?> format, int sstableLevel, long totalSize, ILifecycleTransaction txn, SerializationHeader header) throws IOException
    {
        DiskBoundaries db = cfs.getDiskBoundaries();
        directories = db.directories;
        this.sstableLevel = sstableLevel;
        this.cfs = cfs;
        this.estimatedKeys = estimatedKeys / directories.size();
        this.repairedAt = repairedAt;
        this.pendingRepair = pendingRepair;
        this.isTransient = isTransient;
        this.format = format;
        this.txn = txn;
        this.header = header;
        boundaries = db.positions;
        if (boundaries == null)
        {
            Directories.DataDirectory localDir = cfs.getDirectories().getWriteableLocation(totalSize);
            if (localDir == null)
                throw new IOException(String.format("Insufficient disk space to store %s",
                                                    FBUtilities.prettyPrintMemory(totalSize)));
            Descriptor desc = cfs.newSSTableDescriptor(cfs.getDirectories().getLocationForDisk(localDir), format);
            currentWriter = cfs.createSSTableMultiWriter(desc, estimatedKeys, repairedAt, pendingRepair, isTransient, null, sstableLevel, header, txn);
        }
    }

    private void maybeSwitchWriter(DecoratedKey key)
    {
        if (boundaries == null)
            return;

        boolean switched = false;
        while (currentIndex < 0 || key.compareTo(boundaries.get(currentIndex)) > 0)
        {
            switched = true;
            currentIndex++;
        }

View on GitHub (pinned to 88fd0f6a0e)

Solutions

1. Free disk space or expand the data volumes on the target node, then retry the operation
2. Lower the data footprint (reduce replication, drop unnecessary snapshots, run cleanup/gc) before re-running streaming/repair
3. Add or rebalance data directories in cassandra.yaml so getWriteableLocation can find a volume with sufficient space
4. Check the disk failure policy configuration; a misconfigured policy can mark healthy disks unavailable

Example fix

// mitigation is operational; caller-side guard:
// before
RangeAwareSSTableWriter w = new RangeAwareSSTableWriter(cfs, header, totalSize, ...); // throws
// after
if (cfs.getDirectories().getWriteableLocation(totalSize) == null)
    throw new IOException("Aborting: not enough disk space for " + FBUtilities.prettyPrintMemory(totalSize));
RangeAwareSSTableWriter w = new RangeAwareSSTableWriter(cfs, header, totalSize, ...);

if (cfs.getDirectories().getWriteableLocation(totalSize) == null)
    throw new IOException("Insufficient disk space for " + FBUtilities.prettyPrintMemory(totalSize));

try { writer = new RangeAwareSSTableWriter(cfs, header, totalSize, ...); }
catch (IOException e) {
    if (e.getMessage().startsWith("Insufficient disk space"))
        // free space, then retry or fail the stream gracefully
}

When it happens

Trigger: Creating a RangeAwareSSTableWriter (streaming, repair, or compaction-adjacent writers) when every configured data directory's usable space is below the estimated totalSize, or disk failure policy has blacklisted all volumes.

Common situations: Full disks during streaming/repair in large clusters; undersized test volumes; badly distributed data dirs where all candidate mounts are full.

Understand the failure class

Background: "failed to write file", "Could not save figure", "Error saving remote file" — file write failed: causes and fixes across languages and libraries — this error's family across 38 libraries.

Related errors

• Not enough disk space for stream %s), summaries=%s
• Checksum didn't match (expected: %d, actual: %d)
• FSWriteError (wraps IOException writing sstable component)
• This output stream is in an unsafe state after an asynchrono
• Only {} free across all data volumes. Consider adding more c

AI-assisted analysis of apache/cassandra@88fd0f6a0e (2026-09-10). Data as JSON: /api/errors/98b85f2ea1167218. Report an issue: GitHub.