Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
Hello, Some background for systemd maintainers: ctdb was failing to start with a no such file or drectory error after an upgrade from wheezy to jessie. 2015-01-28 1:07 GMT+01:00 Martin Schwenke mar...@meltin.net: Hi Mathieu, This has stopped failing. While that's obviously excellent, I wish I could explain it. :-( OK. Perhaps a systemd update did something magic? I've only just seen Cameron Norman's suggestion from 2015-01-02 to try systemctl daemon-reload. That would have been a nice idea. Not sure why I didn't get an email from the bug system for Cameron's comment. :-( This was happening on my laptop where reboots are uncommon... so I suppose a reboot could have fixed it too. Yes.I guess. Do packages that provide systemd init need to set some sort of trigger that causes a systemctl daemon-reload? I just tried installing proftpd-basic (random choice of daemon that I've never had installed before), it starts and stops fine and I can't find any evidence of a trigger like that, so that theory seems bogus. :-( systemctl --system daemon-reload is in ctdb postrm but not in postinst (from debhelper). Asking to the systemd maintainers: Is this a known problem? Is this coming from ctdb being an init file before, and a .service now? Answers to questions below... [trimming] thanks for your answers. Cheers, -- Mathieu -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
Am 28.01.2015 um 15:37 schrieb Mathieu Parent: Hello, Some background for systemd maintainers: ctdb was failing to start with a no such file or drectory error after an upgrade from wheezy to jessie. 2015-01-28 1:07 GMT+01:00 Martin Schwenke mar...@meltin.net: Hi Mathieu, This has stopped failing. While that's obviously excellent, I wish I could explain it. :-( OK. Perhaps a systemd update did something magic? I've only just seen Cameron Norman's suggestion from 2015-01-02 to try systemctl daemon-reload. That would have been a nice idea. Not sure why I didn't get an email from the bug system for Cameron's comment. :-( This was happening on my laptop where reboots are uncommon... so I suppose a reboot could have fixed it too. Yes.I guess. Do packages that provide systemd init need to set some sort of trigger that causes a systemctl daemon-reload? I just tried installing proftpd-basic (random choice of daemon that I've never had installed before), it starts and stops fine and I can't find any evidence of a trigger like that, so that theory seems bogus. :-( systemctl --system daemon-reload is in ctdb postrm but not in postinst (from debhelper). Asking to the systemd maintainers: Is this a known problem? Is this coming from ctdb being an init file before, and a .service now? Do you use dh-systemd/init-system-helpers to register your cdtb.service unit? It generates maintainer scripts code to ensure systemctl daemon-reload is called at the correct times. -- Why is it that all of the instruments seeking intelligent life in the universe are pointed away from Earth? signature.asc Description: OpenPGP digital signature
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
Am 28.01.2015 um 15:37 schrieb Mathieu Parent: Hello, Some background for systemd maintainers: ctdb was failing to start with a no such file or drectory error after an upgrade from wheezy to jessie. 2015-01-28 1:07 GMT+01:00 Martin Schwenke mar...@meltin.net: Hi Mathieu, This has stopped failing. While that's obviously excellent, I wish I could explain it. :-( OK. Perhaps a systemd update did something magic? I've only just seen Cameron Norman's suggestion from 2015-01-02 to try systemctl daemon-reload. That would have been a nice idea. Not sure why I didn't get an email from the bug system for Cameron's comment. :-( This was happening on my laptop where reboots are uncommon... so I suppose a reboot could have fixed it too. Yes.I guess. Do packages that provide systemd init need to set some sort of trigger that causes a systemctl daemon-reload? I just tried installing proftpd-basic (random choice of daemon that I've never had installed before), it starts and stops fine and I can't find any evidence of a trigger like that, so that theory seems bogus. :-( systemctl --system daemon-reload is in ctdb postrm but not in postinst (from debhelper). Asking to the systemd maintainers: Is this a known problem? Is this coming from ctdb being an init file before, and a .service now? Do you use dh-systemd/init-system-helpers to register your cdtb.service unit? It generates maintainer scripts code to ensure systemctl daemon-reload is called at the correct times. -- Why is it that all of the instruments seeking intelligent life in the universe are pointed away from Earth? signature.asc Description: OpenPGP digital signature
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
2015-01-28 15:42 GMT+01:00 Michael Biebl bi...@debian.org: Hello Michael, Thanks for your quick response. [...] Do you use dh-systemd/init-system-helpers to register your cdtb.service unit? It generates maintainer scripts code to ensure systemctl daemon-reload is called at the correct times. I use dh_systemd with overrides: http://anonscm.debian.org/cgit/pkg-samba/ctdb.git/tree/debian/rules ctdb.postinst (in the .deb) ends up with only: if [ -x /etc/init.d/ctdb ]; then update-rc.d ctdb defaults /dev/null || exit $? fi and /etc/init.d/ctdb is indeed executable (Is it on your machine Martin?). It seems that update-rc.d calls dameon-reload in some circumstances. Or should ctdb.postinst call it directly? Regards Mathieu Parent -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
Am 28.01.2015 um 15:52 schrieb Mathieu Parent: 2015-01-28 15:42 GMT+01:00 Michael Biebl bi...@debian.org: Do you use dh-systemd/init-system-helpers to register your cdtb.service unit? It generates maintainer scripts code to ensure systemctl daemon-reload is called at the correct times. I use dh_systemd with overrides: http://anonscm.debian.org/cgit/pkg-samba/ctdb.git/tree/debian/rules ctdb.postinst (in the .deb) ends up with only: if [ -x /etc/init.d/ctdb ]; then update-rc.d ctdb defaults /dev/null || exit $? fi and /etc/init.d/ctdb is indeed executable (Is it on your machine Martin?). It seems that update-rc.d calls dameon-reload in some circumstances. Or should ctdb.postinst call it directly? Two issues: 1/ update-rc.d should call daemon-reload for sysv init scripts, it currently doesn't 2/ in your case you ship a native service file, *but* you you install the service file in override_dh_installinit. dh_systemd_enable runs *before* dh_installinit, so the helper doesn't pick up the file and doesn't generate any code for it. My suggestion would be, that you either convince upstream to ship that file upstream (and have it installed by make install) or install it via debian/ctdb.install. dh_install runs before the dh_systemd_* If you fix the installation of the ctdb.service file, the dh-systemd helper should generate maintainer scripts code for that service. Btw, I think you can drop the override_dh_systemd_enable override, it doesn't pass any relevant arguments to dh_systemd, so it seems superfluous. [1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=774799 -- Why is it that all of the instruments seeking intelligent life in the universe are pointed away from Earth? signature.asc Description: OpenPGP digital signature
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
On Wed, 28 Jan 2015 15:52:11 +0100, Mathieu Parent math.par...@gmail.com wrote: I use dh_systemd with overrides: http://anonscm.debian.org/cgit/pkg-samba/ctdb.git/tree/debian/rules ctdb.postinst (in the .deb) ends up with only: if [ -x /etc/init.d/ctdb ]; then update-rc.d ctdb defaults /dev/null || exit $? fi and /etc/init.d/ctdb is indeed executable (Is it on your machine Martin?). Yes, it is executable. peace happiness, martin -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
2015-01-16 9:20 GMT+01:00 Martin Schwenke mar...@meltin.net:
On Tue, 13 Jan 2015 11:26:48 +0100, Mathieu Parent
math.par...@gmail.com wrote:
2014-12-31 23:16 GMT+01:00 Martin Schwenke mar...@meltin.net:
# systemctl start ctdb
Failed to start ctdb.service: Unit ctdb.service failed to load: No such
file or directory.
[...]
Couldn't reproduce here. But I need to test again in a clean VM.
Yikes! :-(
Still not reproducable with:
- a clean VM
- apt-get install ctdb
- sed -i 's/^CTDB_RECOVERY_LOCK/#CTDB_RECOVERY_LOCK/' /etc/default/ctdb
- edit /etc/ctdb/nodes (a one node cluster)
- touch /etc/ctdb/public_addresses
- sudo systemctl start ctdb
- wait ...
- ctdb status - OK (this node)
Anything relevant in /var/log/{syslog,ctdb/log.ctdb}? journalctl -u
ctdb.service?
No, nothing at all. It looks like it is failing somewhere in systemd.
It's /var/log/ctdb/log.ctdb. Have you checked it too?
Well, I do get this from selinux:
Jan 16 18:52:20 rover kernel: audit: type=1107 audit(1421394740.536:41175):
pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0
msg='Unknown permission start for class system exe=/lib/systemd/systemd
sauid=0 hostname=? addr=? terminal=?'
However, I am running in *permissive* mode, so selinux shouldn't be
getting in the way:
# sestatus | grep -i mode
Current mode: permissive
Mode from config file: permissive
My gut is telling me that this has something to do with the selinux
message... but how can this happen in permissive mode?
Was this machine migrated from wheezy? Do you have rsyslog installed?
Yes, this machine was probably upgraded from wheezy. I have rsyslog
installed.
Do you know if there's a way of getting some debug out of systemd?
What journalctl --unit ctdb.service says ?
Is there any red line in systemctl list-units?
Can your try to reboot with sysvinit?
...
Regards
--
Mathieu
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
Hi Mathieu,
This has stopped failing. While that's obviously excellent, I wish I
could explain it. :-(
Perhaps a systemd update did something magic? I've only just seen
Cameron Norman's suggestion from 2015-01-02 to try systemctl
daemon-reload. That would have been a nice idea. Not sure why I
didn't get an email from the bug system for Cameron's comment. :-(
This was happening on my laptop where reboots are uncommon... so I
suppose a reboot could have fixed it too.
Do packages that provide systemd init need to set some sort of trigger
that causes a systemctl daemon-reload? I just tried installing
proftpd-basic (random choice of daemon that I've never had installed
before), it starts and stops fine and I can't find any evidence of a
trigger like that, so that theory seems bogus. :-(
Answers to questions below...
On Tue, 27 Jan 2015 14:51:46 +0100, Mathieu Parent
math.par...@gmail.com wrote:
2015-01-16 9:20 GMT+01:00 Martin Schwenke mar...@meltin.net:
Anything relevant in /var/log/{syslog,ctdb/log.ctdb}? journalctl -u
ctdb.service?
No, nothing at all. It looks like it is failing somewhere in systemd.
It's /var/log/ctdb/log.ctdb. Have you checked it too?
Yes, I had checked it and it was always empty, unless I ran
ctdbd_wrapper directly. That's why I assumed a systemd problem.
Now it logs the expected failure due to broken configuration. I'm
happy with that because I know enough if I want to configure CTDB from
this point... :-)
What journalctl --unit ctdb.service says ?
Is there any red line in systemctl list-units?
Yeah, given that it now works I see the CTDB-specific failure.
Sorry that this was a waste of time and we couldn't work out why it was
happening. :-(
peace happiness,
martin
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
On Tue, 13 Jan 2015 11:26:48 +0100, Mathieu Parent
math.par...@gmail.com wrote:
2014-12-31 23:16 GMT+01:00 Martin Schwenke mar...@meltin.net:
# systemctl start ctdb
Failed to start ctdb.service: Unit ctdb.service failed to load: No such
file or directory.
[...]
Couldn't reproduce here. But I need to test again in a clean VM.
Yikes! :-(
Anything relevant in /var/log/{syslog,ctdb/log.ctdb}? journalctl -u
ctdb.service?
No, nothing at all. It looks like it is failing somewhere in systemd.
Well, I do get this from selinux:
Jan 16 18:52:20 rover kernel: audit: type=1107 audit(1421394740.536:41175):
pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0
msg='Unknown permission start for class system exe=/lib/systemd/systemd
sauid=0 hostname=? addr=? terminal=?'
However, I am running in *permissive* mode, so selinux shouldn't be
getting in the way:
# sestatus | grep -i mode
Current mode: permissive
Mode from config file: permissive
My gut is telling me that this has something to do with the selinux
message... but how can this happen in permissive mode?
Was this machine migrated from wheezy? Do you have rsyslog installed?
Yes, this machine was probably upgraded from wheezy. I have rsyslog
installed.
Do you know if there's a way of getting some debug out of systemd?
peace happiness,
martin
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#774328: [Pkg-samba-maint] Bug#774328: ctdb: Failed to start ctdb.service: Unit ctdb.service failed to load: No such file or directory.
2014-12-31 23:16 GMT+01:00 Martin Schwenke mar...@meltin.net:
Package: ctdb
Version: 2.5.4+debian0-3
Severity: grave
Justification: renders package unusable
Dear Maintainer,
# systemctl start ctdb
Failed to start ctdb.service: Unit ctdb.service failed to load: No such file
or directory.
[...]
Couldn't reproduce here. But I need to test again in a clean VM.
Anything relevant in /var/log/{syslog,ctdb/log.ctdb}? journalctl -u
ctdb.service?
Was this machine migrated from wheezy? Do you have rsyslog installed?
Regards
--
Mathieu
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
