Hundreds of dbus-daemon processes

[Marc Girod]

I find a growing number of root owned dbus-daemon processes.
Eventually, this results in a 'too many processes' error preventing something to fork, e.g. a login.

They get created by batches of 4, every 30 minutes on my box.
My box is a puppet client, so that these 30 minutes are not in a local cron tab.
The pids are 11 units apart (when the box is otherwise idle).
The command line is:
//bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session

~# ps -fu root | grep dbus-daemon | grep -v grep | tail -8
root 17509 1 0 11:25 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
root 17520 1 0 11:25 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
root 17531 1 0 11:25 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
root 17542 1 0 11:25 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
root 17972 1 0 11:55 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
root 17983 1 0 11:55 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
root 17994 1 0 11:55 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
root 18005 1 0 11:55 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session

I reported this first to https://bugs.freedesktop.org/show_bug.cgi?id=74114
and was told to report it to ubuntu instead.
Here were some additional comments:

The symptoms you described probably mean that something running as root is either running /usr/bin/dbus-launch, or attempting to access a D-Bus session via the autolaunch: transport while under a transient X11 session, 4 times every 30 minutes.

If this machine does not intentionally run an X11 GUI or a D-Bus session, uninstalling the dbus-x11 package or making /usr/bin/dbus-launch non-executable might work around this. Don't do that if you use an X11 GUI environment, though.

Since 1.4.8, dbus autolaunch has only been effective when DISPLAY is set and non-empty, so if this is autolaunch, an X11 session must be involved somehow.

ProblemType: Bug
DistroRelease: Ubuntu 12.04
Package: dbus 1.4.18-1ubuntu1.4
ProcVersionSignature: Ubuntu 3.2.0-58.88-generic 3.2.53
Uname: Linux 3.2.0-58-generic x86_64
NonfreeKernelModules: nvidia
ApportVersion: 2.0.1-0ubuntu17.6
Architecture: amd64
Date: Mon Jan 27 14:51:19 2014
MarkForUpload: True
ProcEnviron:
 SHELL=/bin/bash
 PATH=(custom, no user)
 LANG=en_US.UTF-8
SourcePackage: dbus
UpgradeStatus: No upgrade log present (probably fresh install)

[Scott Conger (scott-conger)]

We're also seeing this at my workplace (Salesforce) and people are running scripts to clean up the processes. It's irritating as it can spawn enough processes to prevent you from remotely logging in.

I hit this page when googling around looking for a solution.

[Marc Girod (marc-girod)]

Hi Scott,

I do also work for Salesforce.
It would be an interesting indication if this symptom was *only*
experienced at Salesforce.
It would tell about interactions with specific tools we use.
What could it be?

- puppet
- avahi
...

In fact below is the list of binaries having a pid on my box now.

Now, when I killed my running dbus-daemon processes last time, I got a
crash of compiz:

/usr/bin/compiz
SIGSEGV in_IceTransClose
DistroCodeName: precise
https://bugs.launchpad.net/bugs/969078

with a popup:

Apport
'precise' is no longer under development

Maybe this is an indication?
Thanks
Marc

$ ps -ef | perl -nle 'next if /^UID/;$h{$1}++ if
m%^.{48}\[?([^\s\]/:]+)%;END{print for sort keys %h}'
NetworkManager
acpid
ata_sff
atd
avahi-daemon
bash
bdi-default
bitlbee
bluetooth-applet
cdcwatch
compiz
cpuset
cron
crypto
dbus-daemon
dbus-launch
devfreq_wq
ecryptfs-kthrea
edac-poller
ext4-dio-unwrit
firefox
flush-252
fsnotify_mark
gnome-pty-helper
gnome-screensaver
gnome-session
gnome-terminal
hd-audio0
hd-audio1
iprt
irq
jbd2
kblockd
kdevtmpfs
kdmflush
khelper
khubd
khugepaged
khungtaskd
kintegrityd
kpsmoused
krfcommd
ksmd
ksoftirqd
kswapd0
kthreadd
kthrotld
kworker
lightdm
md
migration
nautilus
netns
nm-applet
nmbd
perl
ps
rpcbind
rsyslogd
scsi_eh_0
scsi_eh_1
scsi_eh_2
scsi_eh_3
scsi_eh_4
scsi_eh_5
scsi_eh_6
sendmail
smbd
ssh
su
sync_supers
telepathy-indicator
udisks-daemon
update-notifier
upstart-socket-bridge
upstart-udev-bridge
watchdog
whoopsie
zeitgeist-datahub

On Tue, Apr 1, 2014 at 12:22 AM, Scott Conger <email address hidden> wrote:
> We're also seeing this at my workplace (Salesforce) and people are
> running scripts to clean up the processes. It's irritating as it can
> spawn enough processes to prevent you from remotely logging in.
>
> I hit this page when googling around looking for a solution.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1273258
>
> Title:
> Hundreds of dbus-daemon processes
>
> Status in "dbus" package in Ubuntu:
> New
>
> Bug description:
> I find a growing number of root owned dbus-daemon processes.
> Eventually, this results in a 'too many processes' error preventing something to fork, e.g. a login.
>
> They get created by batches of 4, every 30 minutes on my box.
> My box is a puppet client, so that these 30 minutes are not in a local cron tab.
> The pids are 11 units apart (when the box is otherwise idle).
> The command line is:
> //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
>
> ~# ps -fu root | grep dbus-daemon | grep -v grep | tail -8
> root 17509 1 0 11:25 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
> root 17520 1 0 11:25 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
> root 17531 1 0 11:25 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
> root 17542 1 0 11:25 ? 00:00:00 //bin/dbus-daemon --fork --print-pid 4 --print-address 6 --session
> root 17972 1 0 11:55 ? 00:00:00 //bin/dbus-daemon --fork --print...

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Avio (aviopene)]

I also have this bug when spawning/killing hundreds of Xpra instances.

[Diego Machado Dias (diegodias-m)]

I have the same bug: a number of /usr/bin/dbus-deamon processes which spawn from "/sbin/init splash".

[OMJ (omjunior)]

I see the same thing on my work machine, an Ubuntu 18.04 and dbus version 1.12.2-1ubuntu1.2.

In my case the processes all look like:
/usr/bin/dbus-daemon --syslog-only --fork --print-pid 5 --print-address 7 --session

If I don't kill the processes an booting the laptop in the morning, by noon I'll have ~4000 processes running.

I regularly kill all dbus-daemons running which have the command line above, and I see no side effects.

I cannot find out which process spawns them, since their PPIDs are all 1, but were probably just adopted by init.

[Pavel Malyshev (afunix)]

Happens on 20.04. Eventually swap is 100% full and machine is rebooted by watchdog due to high (>240) load average.

[Seth Arnold (seth-arnold)]

Pavel, OMJ, maybe execsnoop-bpfcc from bpfcc-tools can help spot what program is starting your dbus-daemons?

Thanks