Arch Linux

thenextdon13

Member

Registered: 2013-01-18

Posts: 70

Another nmbd / systemd problem with network dhcp race condition

Hello-
It seems that there are a lot of these problems going around:
https://bbs.archlinux.org/viewtopic.php?id=151688
https://bbs.archlinux.org/viewtopic.php?id=154159
https://bbs.archlinux.org/viewtopic.php?id=155747

My problem is different but similar

nmbd has failed:

[dylan@storage ~]$ systemctl --failed
UNIT         LOAD   ACTIVE SUB    DESCRIPTION
nmbd.service loaded failed failed Samba NetBIOS name server

LOAD   = Reflects whether the unit definition was properly loaded.
ACTIVE = The high-level unit activation state, i.e. generalization of SUB.
SUB    = The low-level unit activation state, values depend on unit type.

1 loaded units listed. Pass --all to see loaded but inactive units, too.
To show all installed unit files use 'systemctl list-unit-files'.

I am not using NetworkManager, Just dhcpcd service

[dylan@storage ~]$ systemctl list-units | grep -e net -e dhcp
sys-devi...-net-eth0.device loaded active plugged   /sys/devices/pci0000:00/0000:00:15.0/0000:04:00.0/net/eth0
sys-subs...ices-eth0.device loaded active plugged   /sys/subsystem/net/devices/eth0
dhcpcd.service              loaded active running   dhcpcd on all interfaces
network.target              loaded active active    Network

nmbd has clearly failed:

dylan@storage ~]$ sudo systemctl status nmbd.service
nmbd.service - Samba NetBIOS name server
	  Loaded: loaded (/usr/lib/systemd/system/nmbd.service; enabled)
	  Active: failed (Result: exit-code) since Thu 2013-01-17 21:04:10 PST; 4min 51s ago
	 Process: 342 ExecStart=/usr/sbin/nmbd -F (code=exited, status=1/FAILURE)

I can't seem to get any job information from journalctl on it:

[dylan@storage ~]$ journalctl -b -u nmbd
-- Logs begin at Thu 2013-01-17 15:46:04 PST, end at Thu 2013-01-17 21:10:01 PST. --
[dylan@storage ~]$ 

I have the modified service file, that starts up after network:

[dylan@storage samba]$ grep -A 3 Unit /usr/lib/systemd/system/nmbd.service
[Unit]
Description=Samba NetBIOS name server
After=network.target

[dylan@storage samba]$ 

/var/log/samba/log.nmbd on a _working_ boot shows nmbd getting sockets

[dylan@storage samba]$ grep -i -A 10 "sockets 137"  log.nmbd.old 
  Opening sockets 137
[2013/01/17 21:16:02,  5] lib/util_sock.c:165(print_socket_options)
  Socket options:
  	SO_KEEPALIVE = 0
  	SO_REUSEADDR = 1
  	SO_BROADCAST = 1
  	Could not test socket option TCP_NODELAY.
  	Could not test socket option TCP_KEEPCNT.
  	Could not test socket option TCP_KEEPIDLE.
  	Could not test socket option TCP_KEEPINTVL.
  	IPTOS_LOWDELAY = 0
[dylan@storage samba]$ 

/var/log/samba/log.nmbd on a broken boot shows nmbd not getting sockets or binding to an address for some reason.  Perhaps the network card module isn't up yet?

[dylan@storage samba]$ grep -i -A 10 "sockets 137"  log.nmbd
  Opening sockets 137
[2013/01/17 21:18:28,  3] ../lib/util/util_net.c:70(interpret_string_addr_internal)
  interpret_string_addr_internal: getaddrinfo failed for name 0.0.0.0 [Address family for hostname not supported]
[2013/01/17 21:18:28,  0] nmbd/nmbd.c:690(open_sockets)
[dylan@storage samba]$ 

Restarting nmbd makes it start just fine:

[dylan@storage samba]$ sudo systemctl status nmbd.service
nmbd.service - Samba NetBIOS name server
	  Loaded: loaded (/usr/lib/systemd/system/nmbd.service; enabled)
	  Active: failed (Result: exit-code) since Thu 2013-01-17 21:18:28 PST; 16min ago
	Main PID: 346 (code=exited, status=1/FAILURE)
	  CGroup: name=systemd:/system/nmbd.service

[dylan@storage samba]$ sudo systemctl restart nmbd.service
[dylan@storage samba]$ sudo systemctl status nmbd.service
nmbd.service - Samba NetBIOS name server
	  Loaded: loaded (/usr/lib/systemd/system/nmbd.service; enabled)
	  Active: active (running) since Thu 2013-01-17 21:35:14 PST; 6s ago
	Main PID: 1630 (nmbd)
	  CGroup: name=systemd:/system/nmbd.service
		  └─1630 /usr/sbin/nmbd -F

Jan 17 21:35:14 storage systemd[1]: Starting Samba NetBIOS name server...
Jan 17 21:35:14 storage systemd[1]: Started Samba NetBIOS name server.
[dylan@storage samba]$ 

Any suggestions or ideas for troubleshooting where the race condition exists here?
Thanks!

gm25flew

Member

Registered: 2012-12-10

Posts: 4

Re: Another nmbd / systemd problem with network dhcp race condition

Hello,

I am having trouble with nmbd as well.  I can't start it through systemd at all.  I can't even do the restart.  Eventually I get the same errors as you.  I was able to get it running by using interactive mode:

 /usr/sbin/nmbd -i 

Have you solved this yet?
Thanks.

thenextdon13

Member

Registered: 2013-01-18

Posts: 70

Re: Another nmbd / systemd problem with network dhcp race condition

In a sense yes;  I don't reboot.

No, i haven't.  And its really frustrating.  Luckily we have good power here and the box rarely reboots.

I'm not sure if we are sharing the same problem.  My problem has to do with some race condition at startup where nmbd tries to come up before the network is ready, so can't. 

Once the network and server are up, i can restart nmbd with no problem whatsoever.

Sorry i can't be of more help.

colbert

Member

Registered: 2007-12-16

Posts: 809

Re: Another nmbd / systemd problem with network dhcp race condition

Same problem here, it just began happening on my reboot back into Arch from Win7.

└─> sudo systemctl start nmbd
[sudo] password for bobby: 
Job for nmbd.service failed. See 'systemctl status nmbd.service' and 'journalctl -xn' for details.
┌─[ 01/28/13 11:14:34 PM ][ ~ ]
└─> systemctl status nmbd.service
nmbd.service - Samba NetBIOS name server
          Loaded: loaded (/usr/lib/systemd/system/nmbd.service; enabled)
          Active: failed (Result: timeout) since Mon 2013-01-28 04:19:30 EST; 14s ago
         Process: 3277 ExecStart=/usr/sbin/nmbd -F (code=exited, status=0/SUCCESS)

I am able to run the nmbd in interactive mode as gm25flew has kindly pointed out, but it won't work via systemd.

skanky

Member

From: WAIS

Registered: 2009-10-23

Posts: 1,847

Re: Another nmbd / systemd problem with network dhcp race condition

I was having the same issue as gm25flew and colbert this morning, but I've just upgraded samba & smbclient from 3.6.11-2 to 3.6.11-3 and nmbd started on restart.

Last edited by skanky (2013-01-28 09:51:53)

---

"...one cannot be angry when one looks at a penguin."  - John Ruskin
"Life in general is a bit shit, and so too is the internet. And that's all there is." - scepticisle

aesiris

Member

Registered: 2012-02-25

Posts: 97

Re: Another nmbd / systemd problem with network dhcp race condition

In the other thread I saw mention of NetworkManager-wait-online.service
Did you try writing your own "whatever"-wait-online.service?

For example to wait for a global ip I would try:

[Unit]
Description=Whatever Wait Online
Requisite=whatever.service
After=whatever.service
Wants=network.target
Before=network.target

[Service]
Type=oneshot
ExecStart=/usr/bin/bash -c 'timeout=3; for ((i=0;i<timeout;i++)); do [[ $( ip addr show scope global ) ]] && break || sleep 1s; done; (( ($?==0) && (i<timeout) ))'

[Install]
WantedBy=network.target

thenextdon13

Member

Registered: 2013-01-18

Posts: 70

Re: Another nmbd / systemd problem with network dhcp race condition

Even with the newer samba versions, same problem here still:

[dylan@storage ~]$ grep -i sam /var/log/pacman.log  | tail -4
[2012-11-09 07:18] upgraded samba (3.6.8-1 -> 3.6.9-1)
[2012-12-20 09:29] upgraded samba (3.6.9-1 -> 3.6.10-1)
[2013-01-27 21:49] upgraded samba (3.6.10-1 -> 3.6.11-2)
[2013-01-28 14:47] upgraded samba (3.6.11-2 -> 3.6.11-3)
[dylan@storage ~]$ 

and

[dylan@storage ~]$ systemctl --failed
UNIT         LOAD   ACTIVE SUB    DESCRIPTION
nmbd.service loaded failed failed Samba NetBIOS name server

LOAD   = Reflects whether the unit definition was properly loaded.
ACTIVE = The high-level unit activation state, i.e. generalization of SUB.
SUB    = The low-level unit activation state, values depend on unit type.

1 loaded units listed. Pass --all to see loaded but inactive units, too.
To show all installed unit files use 'systemctl list-unit-files'.

As before, a systemctl restart nmbd.service fixes the problem.

gm25flew

Member

Registered: 2012-12-10

Posts: 4

Re: Another nmbd / systemd problem with network dhcp race condition

The updates today fixed it for me.  It started up on it's own without a problem.

thenextdon13

Member

Registered: 2013-01-18

Posts: 70

Re: Another nmbd / systemd problem with network dhcp race condition

@gm25flew

Did you reboot a few times and check each time?  Just curious.  It has worked on/off for me randomly.

Thanks!

thenextdon13

Member

Registered: 2013-01-18

Posts: 70

Re: Another nmbd / systemd problem with network dhcp race condition

@aesiris

I'm not sure how this would help, as its already being specified 'after' network.target.  I will look into your suggestion further, thank you!

I could probably pick some other random service that starts later in the process and add 'After' that service, but that can't be the 'right' way to do it.

I ran into a very similar problem on Ubuntu over a year ago about a year ago, and they provided a fix for their init system:
https://bugs.launchpad.net/ubuntu/+sour … bug/836849

colbert

Member

Registered: 2007-12-16

Posts: 809

Re: Another nmbd / systemd problem with network dhcp race condition

I did the samba packages upgrade, have rebooted and it is working fine now.

donallen

Member

Registered: 2009-11-30

Posts: 23

Re: Another nmbd / systemd problem with network dhcp race condition

I did the samba packages upgrade, have rebooted and it is working fine now.

I have a completely up-to-date x86_64 system, including samba, of course, and nmbd *never* comes up correctly. Ever (since I'm someone who is concerned about the consequences energy waste, I shut my system down at night, so this machine gets rebooted every day). Doing the restart fixes it. Network startup on my machine is done with net-auto-wired.service. I use a static IP address, not dhcp.

This has been going on for months. We need a real fix from the package maintainer, as opposed to users who don't have a full understanding of the issues involved throwing hacks together that work some of the time on some systems.