Windows 11 SMB file copy stalls at 100% for multi-gigabyte transfers to NAS

Environment

Windows 11 22H2 laptops (multiple hardware models), SMB v3 enabled, target NAS: Synology DSM 7.2 on RAID6, clients use wired gigabit Ethernet via an enterprise switch, SMB shares mounted via UNC path \nas\backup

Symptoms

Large file copies (5–50 GB) from local NTFS disk to NAS using File Explorer or robocopy appear to complete, progress hits 100% and transfer dialog closes, but file size on NAS remains smaller than source and transfer stalls without error for 10+ minutes.

Observations

Event Viewer (Microsoft-Windows-SMBClient) shows repeated 'STATUS_CANCELLED' during the final phase; Resource Monitor shows sustained network transmit but no disk I/O on NAS side; robocopy with /V shows '100% - copy may be incomplete' but no explicit error. Wireshark capture on client shows SMB2 WRITE/FLUSH requests with no ACKs after a specific offset. NAS SMART and disk stats normal. Other small files copy fine.

Already attempted

Tried robocopy with /Z and /B (restartable, backup mode) — same behavior. Disabled Windows Defender real-time scanning temporarily — no change. Rebooted NAS and clients. Mapped drive via net use and used xcopy — same stall. Transferred same large files between two Windows machines over same switch successfully. Updated NIC drivers on one client — no effect. Switched to SMB v2 only on client (registry) — transfer completes but is extremely slow.

Question

What diagnostics and reversible configuration changes can identify why large SMB writes reach 100% locally but do not complete on the NAS, and how can I safely finish interrupted copies without data corruption?

Post timeline

COMMUNITY RESPONSES

Verified community troubleshooting

0 replies

No answers yet..

Showing 0 of 0 community responses

AI DIAGNOSTICS

Incident analysis

Updated 1w ago• Trouble ID

Most likely causes

SMB protocol feature mismatch: SMB3 features (multichannel, encryption, signing) interacting poorly with NAS firmware causing write/flush deadlock. 85%

NAS SMB server bug or resource exhaustion handling high-offset writes (particularly with RAID reconstruction or dedupe enabled). 80%

SMB client aggressive write caching or local buffering reports completion before server flush; flush/close not acknowledged and retry logic stalls. 55%

Summary

Symptoms indicate SMB writes progress on the client but the server side stops accepting or acknowledging final writes/flushes. Possible areas: client-side caching/buffering causing apparent completion before server flush, SMB protocol mismatch or opportunistic locking issues, NIC offload or checksum offload interacting with SMB large writes, NAS SMB implementation bug handling high-offset WRITE/FLUSH, or intermediate network device (switch/firmware) dropping specific packet sequences. The fact forcing SMBv2 avoids the hang suggests a protocol/implementation interaction.

Suggested Fixes

Diagnostic step 1

Capture a packet trace on client and NAS during a reproducer (Wireshark/tcpdump) focusing on SMB2/SMB3 WRITE/FLUSH and associated ACKs. Keep captures short to protect privacy; do not share file contents.

Diagnostic step 2

On a test client, disable NIC offloads (Large Send Offload, TCP checksum offload) temporarily: use PowerShell Set-NetAdapterAdvancedProperty or NIC vendor tool, then retry copy to see if behavior changes.

Diagnostic step 3

Temporarily disable SMB caching on client by setting DisableFileSync and related registry keys for SMB to force synchronous writes; test small reproducer files first.