Update with LogLevel=debug enabled, which captured the actual failure (attached: full trace). The Protocol error is misleading — it's downstream of the real failure. The debug trace shows: (sd-gens)[690658]: (sd-exec-strv) terminated by signal ALRM. systemd[1]: (sd-gens) failed with exit status 1. systemd[1]: Failed to fork off sandboxing environment for executing generators: Protocol error systemd[1]: Freezing execution. Generators began executing at 01:00:48 and the SIGALRM fired at 01:02:18 — exactly 90 seconds later, i.e. the generator timeout. A system generator did not complete within the 90s budget, was killed by the alarm, and that caused (sd-gens) to fail and PID1 to freeze. In the trace, the generators that touch NFS (systemd-fstab-generator parsing NFS entries in /etc/fstab, rpc-pipefs-generator) do not log a "succeeded" line before the alarm, whereas the non-NFS generators (systemd-debug-generator, cloud-init-generator, systemd-sysv-generator, snapd-generator, etc.) all complete. This points to a generator blocking on a hard NFSv4 mount during the re-exec. Trigger was a daemon-reexec from a session scope (Reexecuting requested from client PID ... (unit session-NNNNN.scope)), not the apt timer — so any re-exec source can trigger it, not only package upgrades. Mitigation that resolves it: converting the NFS /etc/fstab entries to noauto,x-systemd.automount (keeping hard), which removes the synchronous NFS access from the generator path during re-exec. After this change, a deliberate systemctl daemon-reexec survives where it previously froze. Not yet root-caused: why the generator stalls. The same NFS volume is mounted on other hosts without issue, so it does not appear to be the storage itself — more likely the interaction of this host's hard+fstab mount being probed synchronously by the fstab generator under the 90s budget during re-exec, possibly coinciding with transient server latency. Open question whether systemd should isolate generators from blocking on network filesystems during re-exec so a single slow generator can't freeze PID1. Environment: Ubuntu 24.04, systemd 255.4-1ubuntu8.16, VMware guest, two NFSv4 hard mounts. Reproduced 2026-06-03, 2026-06-09, 2026-06-22.
** Attachment added: "freeze-rootcause-2026-06-22.txt" https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/2156232/+attachment/5978456/+files/freeze-rootcause-2026-06-22.txt -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2156232 Title: systemd 255.4-1ubuntu8.16: PID1 freezes on daemon-reexec after systemd point-release upgrade (generator sandbox fork fails, EPROTO) To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/systemd/+bug/2156232/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
