Re: Process kill on logout

2017-06-30 Thread davidson

On 26 Jun 2017, Alan wrote:


I have just upgraded from Jessie to Stretch and my arrangement where
I logout and then mythtv shuts my computer down and sets a wakeup
time just before the next recording time has stopped working.


I have no help to offer regarding mythtv, or your shutdown/wakeup
scheme.

I'm interested in your GNU screen problems, though:


In the past when that happens I have set up "screen", and entered an
used "sleep 1m && who -u" to find out why its not detecting me as
logged out.  I set that command running, detach from screen, logout,
ensure the minute is up, log back in and re-enter my screen session
to see what it made of things.

That no longer works.


So, just to be sure we are on the same page, you do all of the above
as userjoe and then, after waiting 60 seconds and logging back in
again as userjoe,

 $ screen -ls

tells you that there are no sockets in /run/screen/S-userjoe ?


It appears that systemd now kills off all processess associated with
the user - or rather I found a bug report that said that had
happened about a year ago as the default had changed.


You are refering to one of these bug reports, I guess, which were
marked fixed in version systemd/230-2 ?

 #825941 - screen is killed by systemd when the user logs out
 https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=825941

 #825394 - systemd kill background processes after user logs out
 https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=825394

I have upgraded a minimal jessie install to stretch, and installed GNU
screen after the upgrade, and I am unable to replicate the breakage. I
log out, and my detached screen sessions survive.

It would be interesting to know what makes the difference.

What output do you get for the following four commands? (I include
output from my own system here for comparison.)

 $ systemd --version
 systemd 232
 +PAM +AUDIT +SELINUX +IMA +APPARMOR +SMACK +SYSVINIT +UTMP
 +LIBCRYPTSETUP +GCRYPT +GNUTLS +ACL +XZ +LZ4 +SECCOMP +BLKID
 +ELFUTILS +KMOD +IDN

 $ screen -v
 Screen version 4.05.00 (GNU) 10-Dec-16

 $ uname -a
 Linux magicbucket 4.9.0-3-amd64 #1 SMP Debian 4.9.30-2+deb9u2 (2017-06-26) 
x86_64 GNU/Linux

 $ grep '^[^#]' /etc/apt/sources.list{,.d/*}

 /etc/apt/sources.list:deb http://debian.univ-lorraine.fr/debian/ stretch main
 /etc/apt/sources.list:deb-src http://debian.univ-lorraine.fr/debian/ stretch 
main
 /etc/apt/sources.list:deb http://security.debian.org/ stretch/updates main
 /etc/apt/sources.list:deb-src http://security.debian.org/ stretch/updates main
 /etc/apt/sources.list:deb http://debian.univ-lorraine.fr/debian/ 
stable-updates main
 /etc/apt/sources.list:deb-src http://debian.univ-lorraine.fr/debian/ 
stable-updates main
 grep: /etc/apt/sources.list.d/*: No such file or directory



Process kill on logout

2017-06-25 Thread Alan Chandler
I have just upgraded from Jessie to Stretch and my arrangement where I 
logout and then mythtv shuts my computer down and sets a wakeup time 
just before the next recording time has stopped working.  In the past 
when that happens I have set up "screen", and entered an used "sleep 1m 
&& who -u" to find out why its not detecting me as logged out.  I set 
that command running, detach from screen, logout, ensure the minute is 
up, log back in and re-enter my screen session to see what it made of 
things.


That no longer works.  It appears that systemd now kills off all 
processess associated with the user - or rather I found a bug report 
that said that had happened about a year ago as the default had changed. 
That bug report said that /etc/systemd/logind.conf should allow you to 
change that default behaviour by setting KillUserProcesses=no.



Firstly, although commented out, that setting is already showing "no", 
which is supposed to mean that this is the default value anyway.  
Secondly setting it explicitly to "no" doesn't work, screen sessions 
still appear to have been wiped out by the act of logging out.


Can anyone tell me what to do to allow screen sessions to survive logout.




--
Alan Chandler
http://www.chandlerfamily.org.uk