The result being the following: Mar 04 11:26:05 tau sshd[442]: error: Bind to port 2243 on 10.230.131.4 failed: Cannot assign requested address. Mar 04 11:26:05 tau systemd[1]: sshd.service: Main process exited, code=exited, status=255/EXCEPTION Mar 04 11:26:05 tau sshd[442]: fatal: Cannot bind any address. Mar 04 11:26:05 tau systemd[1]: sshd.service: Failed with result 'exit-code'. It attempts to bind to an ip address before it is actually configured and available. There is an easy fix for that. It is to use the systemd target network-online.target instead of the target network.target. Details are provided here: https://www.freedesktop.org/wiki/Software/systemd/NetworkTarget/ I did fix your service unit file very easily by changing it to: [Unit] Description=OpenSSH Daemon Wants=sshdgenkeys.service network-online.target After=sshdgenkeys.service network-online.target [Service] ExecStart=/usr/bin/sshd -D ExecReload=/bin/kill -HUP $MAINPID KillMode=process Restart=always [Install] WantedBy=multi-user.target
As a side note, I believe that I have stumbled into that bug because I specify a specific ip address of my multi-homed server in sshd_config at the field ListenAddress. as denoted at the bottom of https://www.freedesktop.org/wiki/Software/systemd/NetworkTarget/ servers using the address 0.0.0.0 are unaffected. Perhaps that an elegant solution to not affect the boot time of systems that don't have this issue could be to bundle 2 service units with openssh. 1. The current one that wants network.target 2. A new unit that wants network-online.target for the sshd instances that listen to specific addresses only available when the network is online.
Here is one last idea. Considering how vital sshd is to admin a headless system, I think that it could be a good idea to never giveup attempting to restart it... This could be done with this: [Service] ExecStart=/usr/bin/sshd -D ExecReload=/bin/kill -HUP $MAINPID KillMode=process Restart=always StartLimitIntervalSec=0 RestartSec=20s
The OpenSSH team does not supply a systemd unit file so there's nothing we can do about this. You will need to report this to your OS vendor.
sorry about that. I'll report to the appropriate recipient. thank you nonetheless for the fast reply.
Good luck! Feel free to add a link here to the vendor report in case anyone else is having a similar issue and finds this bug.
closing resolved bugs as of 8.6p1 release