How to Fix ‘Transport Endpoint Is Not Connected’ in Linux

Navigating directories, listing files, or executing read/write operations on a mounted partition in Linux only to encounter the error “Transport endpoint is not connected” (POSIX error code ENOTCONN, errno 107) halts shell workflows immediately. Unlike standard filesystem errors that indicate missing files or permission denials, this specific error indicates that the kernel Virtual File System (VFS) still holds an active mount point descriptor in its directory table, but the userspace daemon or network transport backing that filesystem has terminated abnormally, crashed, or closed its communication socket. Most commonly seen with Filesystem in Userspace (FUSE) implementations—such as SSHFS, GlusterFS, Ceph, Rclone, or Google Cloud Storage FUSE—the mountpoint enters a broken zombie state where any directory traversal generates ENOTCONN. This technical troubleshooting guide explains why transport endpoints disconnect and details safe command-line methods to unmount and restore the mountpoint.

What Does ‘Transport Endpoint Is Not Connected’ Mean?

The Linux VFS routes filesystem system calls (such as stat(), open(), or readdir()) to the driver responsible for that mount point. The error occurs when:

  • Crashed FUSE Userspace Process: When using tools like sshfs or rclone, the underlying background daemon crashes or is terminated by Linux’s Out-Of-Memory (OOM) killer. The kernel still sees the directory as mounted, but there is no process listening on the FUSE socket.
  • Dropped Network Transport Socket: Remote network filesystems (NFS, GlusterFS, or SSHFS) experiencing TCP socket resets or severe network dropouts where the connection terminates without a clean TCP FIN/RST handshake.
  • Deadlocked File Descriptors: A process holding an exclusive file lock inside the mountpoint while the server-side host reboots.
  • Improperly Terminated Background Mounts: Unmounting a filesystem using standard commands while child threads are actively reading files.

Method 1: Perform a Lazy Unmount Using umount -l

When a transport endpoint crashes, running standard umount /path/to/mount usually fails with “device is busy” or returns the exact same transport error. A lazy unmount cleanly detaches the filesystem from the directory hierarchy immediately.

  1. Open your Terminal prompt.
  2. Execute a lazy unmount on the broken directory:
    sudo umount -l /path/to/mountpoint
    The -l (lazy) flag detaches the filesystem from the directory tree immediately, cleaning up all references as soon as active read/write operations cease.
  3. If a lazy unmount still encounters resistance, attempt a forced unmount:
    sudo umount -f /path/to/mountpoint
  4. Confirm that the mountpoint is cleanly detached by running:
    df -h | grep /path/to/mountpoint

Method 2: Unmount via fusermount -u

If the disconnected mount is a user-level FUSE mount (such as SSHFS or Rclone), using FUSE’s native unmount utility cleans the descriptor safely without requiring root privileges.

  1. Run fusermount with the unmount parameter:
    fusermount -u /path/to/mountpoint
  2. If the endpoint reports that it is busy, use the lazy fusermount option:
    fusermount -uz /path/to/mountpoint
    The -z option corresponds to lazy unmounting in FUSE.
  3. Verify that the directory returns to a standard empty local directory:
    ls -la /path/to/mountpoint

Method 3: Terminate Orphaned FUSE and SSHFS Processes

If an orphaned background daemon is still holding half-open communication pipes, terminating the zombie daemon is required before remounting.

  1. Search for any lingering background processes associated with the mount:
    ps aux | grep -E "sshfs|rclone|glusterfs"
  2. Identify the Process ID (PID) of the hanging process.
  3. Terminate the lingering process:
    sudo kill -9 PID
  4. Unmount the mountpoint again using fusermount -uz /path/to/mountpoint.

Method 4: Restart Remote Services (NFS and GlusterFS)

For enterprise clustered storage volumes, restarting the client storage daemon restores socket connectivity.

  1. For GlusterFS deployments:
    sudo systemctl restart glusterd
  2. For NFS client mounts:
    sudo systemctl restart nfs-common
    sudo systemctl restart rpcbind
  3. Once the services are active, re-evaluate remote export visibility:
    showmount -e remote_server_ip

Method 5: Reconnect the Filesystem with Keepalive Options

When remounting network filesystems like SSHFS, configuring automated TCP keepalive packets prevents firewalls and NAT routers from dropping idle sockets.

  1. Remount your remote directory using recommended stability flags:
    sshfs user@remote_host:/remote/path /local/mountpoint -o reconnect,ServerAliveInterval=15,ServerAliveCountMax=3
    These flags instruct SSHFS to automatically re-establish dropped sockets and send keepalive heartbeats every 15 seconds.
  2. For Rclone cloud mounts, add the daemon buffer parameters:
    rclone mount remote:bucket /local/mountpoint --daemon --vfs-cache-mode writes --timeout 1m
  3. Test writing a temporary file to verify that the transport channel is active and stable:
    touch /local/mountpoint/test.tmp && rm /local/mountpoint/test.tmp

Get the best tech tips delivered straight to your inbox.

Join thousands of readers mastering Apple, Google, Microsoft, and Linux.