Confusing log message "init: failsafe main process (...) killed by TERM signal"

[Removed by request]

I'm using Ubuntu 13.04 dev with upstart 1.6.1-1ubuntu1. On booting my system I'm seeing in /var/log/kern.log the line "Dec 18 18:17:46 ubuntu kernel: <12>[ 19.939989] init: failsafe main process (1033) killed by TERM signal".

[Karma Dorje (taaroa)]

$ dmesg | grep kill
[ 17.093010] init: failsafe main process (961) killed by TERM signal

not a bug → invalid

[Removed by request (removed3425744)]

Can you give some more details why this is not a bug?

[Mohammad Shamma (mohammadshamma)]

I second Sworddragon's question.

[James Hunt (jamesodhunt)]

 'failsafe' is a job that will force the boot to proceed should that be required.

The job will always start but in the normal case (where no problems are detected) it will be stopped automatically using SIGTERM as shown by the message.

[anatoly techtonik (techtonik)]

I am confused and it is a bug.

[James Hunt (jamesodhunt)]

This is not a bug: the failsafe job detects if the system failed to boot and takes appropriate action. However, in all likelihood your system will boot as expected so the failsafe job (which is always started) is stopped. Upstart does this by sending the TERM signal, hence the message which can be safely ignored.

[Removed by request (removed3425744)]

I think the most annoying thing is that this causes an entry in dmesg. Wouldn't it be possible that failsafe can exit itself with 0 if it detects that the system has booted successfully instead of waiting for a TERM signal? For example instead of sending the TERM signal Upstart could emit an own event for this which failsafe also listens on the "stop on" stanza. Or doesn't this work?

[Volker Siegel (vsiegel)]

I understand that the process reporting the line is working just fine;
But I think that, in itself, showing the message in the log is a bug;
It is bound to be very distracting when reading the log after something went wrong, trying to spot suspicious messages.

Without background, I read it as saying that the init process tells me that some important ("main") process was killed,
even though it should be failsafe.
Now, reading the comments above, it actually tells me that boot went fine.

I think it can be solved by changing the message to be less ambiguous and ideally to indicate it's not indicating failure of anything.

Just looking at the message, it seems to be used in other places too - so it would be impossible to just replace the whole text cleanly. But it could be good enough to make minor changes only, ie making more clear that it's not the main process that is failsafe::

Original:
"init: failsafe main process killed by TERM signal"

proposed change:
"init: job "failsafe" - main process killed by TERM signal"
or
"init: "failsafe": main process killed by TERM signal"

Could a change like this work for the general message in other cases too?

[atimonin (atimonin)]

I was also confused with this "main process killed by TERM".
And I also shure the message should be more descriptive about "killed, but it is OK"

[Erik Lovlie (eriklovlie)]

I came here because of the same reason the original reporter: the dmesg entry is confusing. Hence this bug should not be "invalid" now that the title has been changed. It might be low priority but it is not invalid.

[robotarmy (ry-white)]

Yep, same as Erik