Direct answer

Backups, antivirus scans, panel jobs, log compressors and other services can create lag spikes. Correlate their start times with spark data before disabling anything.

Correlate the activity

  1. Capture /spark health and /spark profiler start --timeout 120 during the slowdown.
  2. On Linux, read top or ps -eo pid,comm,%cpu,%mem --sort=-%cpu | head for CPU consumers; use vmstat 1 for system pressure.
  3. Check scheduled backups, security scans, panel maintenance and disk compression jobs around the same timestamp.
  4. For a systemd unit, journalctl -u minecraft.service --since "1 hour ago" can help align game restart or crash times.

Reduce interference without weakening safety

Move compression and backups to a lower-traffic window or another host where feasible. Prefer a validated backup schedule to simply stopping backups. Do not disable security software or kill unknown processes merely because CPU use is high. Hosting providers may run tasks outside your VM that you cannot inspect.

Version and safety notes

Use the management panel or operating-system tooling applicable to your server. Do not stop unrelated services or security processes just to free resources.

Prove the fix

Compare MSPT spikes before, during and after one scheduled task. Repeat the test after moving only that job; if the lag remains, investigate the next bottleneck rather than assuming the first process was responsible.

References

Check documentation for the versions and hosting environment you actually run.