Development
Unused NFS Stopped My Orange Pi 3 LTS Home Assistant Server From Shutting Down

The short answer: My Orange Pi 3 LTS running Home Assistant would not finish shutting down or rebooting. Logs pointed toward the installed NFS server, although they did not capture the exact failure. This host did not use NFS, so I simulated removal, purged its four NFS-related packages, and verified Home Assistant still ran. I then tested shutdown and reboot myself, and both worked.
This is a record of one machine, not a blanket recommendation to remove NFS. If your host serves NFS exports, mounts NFS shares, or gives Docker containers NFS-backed storage, check those dependencies before changing packages.
The Orange Pi and the misleading reboot
The machine is an Orange Pi 3 LTS with four ARM Cortex-A53 cores and 1.9 GiB of reported RAM. It runs Debian GNU/Linux 12 (bookworm) on arm64 with kernel 5.16.17-sun50iw6. Home Assistant and Home Assistant Supervisor run in Docker. This is a Debian and Docker host, not Home Assistant OS.
Its root filesystem is /dev/mmcblk2p1 (7.2 GB), with /mnt/hassio on /dev/mmcblk0p1 (10 GB) and /mnt/docker on /dev/mmcblk0p2 (49.5 GB). Those device names describe the observed layout; I did not use them to identify the physical storage media.
Running shutdown -h now did not complete. Reboot appeared stuck too. The fact that the machine came back later was misleading: before the fix, those later boots followed a physical power off and power on. A new boot timestamp alone would have been poor evidence that a software reboot had succeeded.
What the logs showed
I connected over SSH and checked systemd and historical syslog. On September 29, 2026, in the +08:00 timezone, two shutdown attempts included these entries:
00:35:16 systemd[1]: Stopping nfs-server.service - NFS server and services...
06:35:35 systemd[1]: Stopping nfs-server.service - NFS server and services...
I found no matching Stopped nfs-server.service entry for either attempt before the next boot. That made NFS a strong lead, but the final visible log line does not establish exactly where shutdown blocked. I did not capture a systemd or kernel trace that proves a particular deadlock.
Before removal, nfs-server and rpcbind were active, and ports 2049 and 111 were listening. That established that NFS components were running; it did not establish that Home Assistant needed them.
Check whether NFS is actually in use
I inspected exports, mounts, Docker storage, and the host's listening ports. /etc/exports contained only comments, exportfs -v returned no active exports, there were no NFS client mounts, and no Home Assistant Docker volume used NFS. Home Assistant's storage was locally bind mounted. This dedicated host had no reason to serve or mount NFS.
These are useful checks on a similar Debian host. Review the output and your own Docker Compose or container configuration before deciding anything is unused:
systemctl status nfs-server rpcbind
cat /etc/exports
sudo exportfs -v
findmnt -t nfs,nfs4
sudo ss -lntup | grep -E ':(111|2049)\b'
docker inspect homeassistant hassio_supervisor
A listening NFS or RPC port shows a service is active, not that it has clients. Conversely, an empty export list does not rule out NFS client mounts or container volume dependencies. If any of those checks reveal real use, do not follow my purge command without planning a replacement for that use.
Simulate the purge, then remove the unused packages
Before changing the machine, I ran an apt simulation and inspected the proposed removals. The simulation showed only the four NFS-related packages below, not Docker or Home Assistant:
sudo apt-get -s purge nfs-kernel-server nfs-common rpcbind libnfsidmap1
On September 29, 2026 at 08:50 +08:00, I ran the actual purge:
sudo env DEBIAN_FRONTEND=noninteractive apt-get purge -y \
nfs-kernel-server nfs-common rpcbind libnfsidmap1
The removed versions were nfs-kernel-server and nfs-common 1:2.6.2-4+deb12u1, rpcbind 1.2.6-6+b1, and libnfsidmap1 1:2.6.2-4+deb12u1. Do not copy this command onto a host that depends on NFS. In particular, nfs-common can matter to a machine that mounts remote NFS shares even if it exports nothing.
Verify the service and Home Assistant afterward
After the purge, I confirmed the four packages were gone, NFS and RPC services were inactive, nothing listened on TCP or UDP port 2049 or 111, and there were no NFS mounts. The homeassistant and hassio_supervisor containers were still running. After a later boot, systemctl --failed showed no failed units.
For a similar check, inspect the actual package and service state rather than assuming a successful apt exit proves the application is healthy:
dpkg-query -W nfs-kernel-server nfs-common rpcbind libnfsidmap1
systemctl status nfs-server rpcbind
findmnt -t nfs,nfs4
sudo ss -lntup | grep -E ':(111|2049)\b'
docker ps --filter name=homeassistant --filter name=hassio_supervisor
systemctl --failed
A nonzero result from dpkg-query or grep can be expected when a package or listener is absent; read their output rather than treating those commands as a combined pass or fail script. Check the Home Assistant interface as well as the container state on your own host.
I subsequently tested shutdown and reboot, and both completed normally. That before and after result makes removal of unused NFS the practical fix on this machine. An unrelated system package upgrade happened earlier the same day, so I cannot prove that the purge alone explains every part of the failure. The exact low-level shutdown mechanism remains unknown.
If you are troubleshooting a similar host, the useful sequence is to preserve the logs, distinguish a software reboot from a physical power cycle, prove whether NFS is used, simulate any package removal, and verify both Home Assistant and shutdown afterward. More Linux and development notes are in the blog archive.