Direct answer

A server may fail writes when filesystem capacity or inodes are exhausted. Slow storage throughput and full disks are different faults; diagnose both rather than deleting files at random.

How to check it

  1. From the actual server directory, run df -h . and df -i . to check available bytes and inodes for that mounted filesystem.
  2. Run du -sh world* logs plugins only on paths that exist to identify large consumers; inspect backups in their actual storage location.
  3. Review logs for write failures, region I/O errors, full filesystem warnings or long save times.

What to change

cd /path/to/minecraft-server
df -h .
df -i .
du -sh world* logs plugins 2>/dev/null
  1. Keep a verified external backup before any cleanup; archive or rotate expendable old logs/archives deliberately.
  2. Free space on the filesystem holding the world, not an unrelated mount.
  3. If space and inodes are healthy, measure latency/IOPS separately instead of assuming the disk is full.

Software and version notes

Compatibility and safety

World data, logs, backups and plugin databases can reside on different mounts. df reports the filesystem containing the specified path and df -i checks inode counts; do not delete live world/region data.

Verify the result

Verify new chunk saves and a test backup complete successfully; compare I/O wait and save durations to the baseline.

References