respawn.log spammed with errors when there is no serial port on the machine

[janeczku]

RancherOS Version: (ros os version)
v1.4.0
Where are you running RancherOS? (docker-machine, AWS, GCE, baremetal, etc.)
vSphere

On hosts that don't one or two serial ports connected, the following errors are logged in /var/log/boot/respawn.log every 10 seconds.

$ cat respawn.log
time="2018-09-07T19:11:28Z" level=debug msg="START: [respawn -f /etc/respawn.conf] in /"
time="2018-09-07T19:11:28Z" level=info msg="respawn RancherOS
built: '2018-09-07T17:58:59Z', 289ff05"
time="2018-09-07T19:11:38Z" level=error msg="/sbin/agetty -n -l /usr/bin/autologin -o rancher:ttyS0 ttyS0 : exit status 1"
time="2018-09-07T19:11:38Z" level=error msg="/sbin/agetty -n -l /usr/bin/autologin -o rancher:ttyS1 ttyS1 : exit status 1"
(...)

The ISO and many of the other images enable autologin for one or multiple serial ports by default.

Possible fix: Check if the serial port device is actually connected to a hardware device before adding to respawn.conf.

os/cmd/control/console_init.go

Line 244 in 6a676a9

if allowAutoLogin && strings.Contains(cmdline, fmt.Sprintf("rancher.autologin=%s", tty)) {

[niusmallnan]

If users have installed ros to the disk via ros install, there will be no problem, because of the rancher.autologin items will be removed.

If users boot an instance via docker-machine, they will hit this issue.
This is because ttyS0 and ttyS1 are not a tty sometimes, we can check the syslog:

$ tail -f /var/log/syslog
Oct 18 05:22:59 test-respawn agetty[1764]: /dev/ttyS1: not a tty
Oct 18 05:23:09 test-respawn agetty[1765]: /dev/ttyS0: not a tty
Oct 18 05:23:09 test-respawn agetty[1766]: /dev/ttyS1: not a tty
Oct 18 05:23:19 test-respawn agetty[1767]: /dev/ttyS0: not a tty
Oct 18 05:23:19 test-respawn agetty[1768]: /dev/ttyS1: not a tty

[rootwuj]

Version - rancher/os:v1.5.0-rc1 11/30
Verified fixed