init.d scripts

[snacknuts]

Today I noticed that all /etc/init.d/ scripts that "--exec" shell scripts have started failing.

for example:

/etc/init.d/courier-imap start
* Starting courier-imapd ... [ !! ]

/etc/init.d/jabberd and courier-imap-ssl also fail.

syslog indicates that the shell scripts being exec'd start and are then immediatly stopped (killed?)

please help

[jk3us]

baselayout has been updated several times lately, did you do an etc-update after your last emerge world?

[snacknuts]

Figured out the problem. runscript doesn't like running shell scripts via start-stop-daemon. my solution was to rewrite the startup scripts and not use start-stop-daemon

[timotheus25]

Today I noticed that all /etc/init.d/ scripts that "--exec" shell scripts have started failing.

This is one royal pain. Is it the intended behavior on the part of the devs that only start-stop-daemon runs binaries? What about chuid for scripts? What is the best approach to compensate? Was this caveat discussed in Gentoo Weekly Newsletter or something comparable?

See Bug 143951 , it helps for the slight case of having the shell script change its process name.

[UberLord]

This is one royal pain. Is it the intended behavior on the part of the devs that only start-stop-daemon runs binaries?

Technically you can run scripts - provided that the script is the daemon and uses #!/bin/bash instead of #!env bash
The problem is that what s-s-d executes is expected to be the daemon. In courier-imaps case what is executes is really a script that launches daemons. I submitted fixed scripts for that over a year ago, but for various reasons (devs retired is one) i never got stabled. That is now happening

Same for jabber I think as it was #!env python instead of #!/usr/bin/python

What about chuid for scripts? What is the best approach to compensate?

Again, you can use chuid for scripts provided they meet the above criteria - or move the logic of the "script" into the init script and solve the issue that way.

Was this caveat discussed in Gentoo Weekly Newsletter or something comparable?

Actually yes. I sent out 3 e-mails to the gentoo-dev mailing list, and no-one objected. I also emailed how to solve existing init scripts. I informed people though the correct channels.

[timotheus25]

Technically you can run scripts - provided that the script is the daemon and uses #!/bin/bash instead of #!env bash
The problem is that what s-s-d executes is expected to be the daemon.

Thanks, but I am still looking for a solution.
Jabber is a good example of a now broken service. I have jabberd-2.0.11 installed (latest).

Code: Select all

% head -n 7 /usr/bin/jabberd 
#!/bin/sh
#-*-Perl-*-

exec perl -w -x $0 "$@"

#!perl

Code: Select all

% grep -A 2 -- --start /etc/init.d/jabber
        start-stop-daemon --start -c jabber:jabber \
                --make-pidfile --pidfile /var/run/jabber/jabberd.pid \
                --background -q -x /usr/bin/jabberd

Unfortunately, this does not work. Jabberd runs perfectly for one second, but then s-s-d kills it and fails the service.
Also, I want to run many python scripts that create PID files but do not rename their process as a daemon. What can I do?

For example:

Code: Select all

% cat /usr/local/jabber/bin/pyaim-t.sh
#!/bin/bash
CMD=/usr/lib/python2.4/site-packages/pyaim-t/pyaim-t.py
exec python $CMD -c /etc/jabber/pyaim-t.xml -l /var/log/jabber/pyaim-t.log

Code: Select all

% grep -A 3 -- --start /etc/init.d/pyaim-t
        start-stop-daemon --name pyaim-t.sh --start --background --chuid jabber:jabber \
                --exec /usr/local/jabber/bin/pyaim-t.sh
        true
        eend $?

And the change of the $? is necessary for it to pass the 'start' function.
Since all of these create PID files, stopping with s-s-d is obvious.

Would bugging baselayout for a a s-s-d man-page example w.r.t. scripts be reasonable?

[UberLord]

Technically you can run scripts - provided that the script is the daemon and uses #!/bin/bash instead of #!env bash
The problem is that what s-s-d executes is expected to be the daemon.

Thanks, but I am still looking for a solution.
Jabber is a good example of a now broken service. I have jabberd-2.0.11 installed (latest).

jabber is a bad case for start-stop-daemon. The last time I looked at the script which is the "daemon" it really launched 4 separate daemons.
http://bugs.gentoo.org/show_bug.cgi?id=102546

Also, I want to run many python scripts that create PID files but do not rename their process as a daemon. What can I do?

For example:

Code: Select all

% cat /usr/local/jabber/bin/pyaim-t.sh
#!/bin/bash
CMD=/usr/lib/python2.4/site-packages/pyaim-t/pyaim-t.py
exec python $CMD -c /etc/jabber/pyaim-t.xml -l /var/log/jabber/pyaim-t.log

Code: Select all

% grep -A 3 -- --start /etc/init.d/pyaim-t
        start-stop-daemon --name pyaim-t.sh --start --background --chuid jabber:jabber \
                --exec /usr/local/jabber/bin/pyaim-t.sh
        true
        eend $?

It's called start-stop-daemon for a reason - and what you are trying to run is NOT a daemon.
This may work

Code: Select all

start-stop-daemon --start --exec python -- background --chuid jabber:jabber \
   -- /usr/lib/python2.4/site-packages/pyaim-t/pyaim-t.py \
   -c /etc/jabber/pyaim-t.xml -l /var/log/jabber/pyaim-t.log

Would bugging baselayout for a a s-s-d man-page example w.r.t. scripts be reasonable?

Sure - make a patch for our man page and we'll include it

[halfgaar]

I submitted fixed scripts for that over a year ago, but for various reasons (devs retired is one) i never got stabled. That is now happening

Does that mean the next update of courier-imap has this fixed? I mean, I can hack startup scripts myself, but on the next update I overwrite this init.d script, so if there is no official solution, I have to remember not to overwrite the script on etc-update. That I remember is unlikely to happen, because one shouldn't make custom modification to init.d scripts, and therefore I always overwrite them with etc-update.

[UberLord]

I submitted fixed scripts for that over a year ago, but for various reasons (devs retired is one) i never got stabled. That is now happening

Does that mean the next update of courier-imap has this fixed?

I believe that version 4.0.4 in portage right now has fixed scripts.
http://bugs.gentoo.org/show_bug.cgi?id=143830

[optics]

vmware startup is suffering from a similar problem after baselayout update. On boot it seems to fly past vmware start (maybe a parallel start?)

When in terminal I notice none of the vmware eth adapters are up. I try to manually start the vmware init.d script and it simply hangs. Is a reinstall of vmware necessary after such a baselayout change?

I'm also not sure what is the best way to look for errors. The usual messages that appeared on individual components of larger init.d starts are not there (ok's or !!'s). I also ran metalog and don't see any activity in /var/log/everything.

[optics]

interesting.. a reemerge fixed vmware.

still have some funkiness with this new baselayout.. shutdown doesn't complete hangs on various services..

[zaphyr]

same here
shutdown stops after hostname, have to use ctrl+c to kill it and then it says it can not stop net.lo and some other stuff.
after boot, I have noticed that samba and cupsd is not running, maybe others as well.
example:

Code: Select all

# /etc/init.d/cupsd status
 * status:  starting

no fix for this?

[zaphyr]

disabling parallel startup fixed my problems

[lyallp]

I had slightly different problems with my startup scripts which you can look at http://forums.gentoo.org/viewtopic-t-49 ... ight-.html

I had a custom written daemon script (popfile to be precise) which was based upon other scripts.

Maybe my resolution might help others (see last entry of the post).