37

I get the following pieces of information at the startup, takes about 3 to 5 minutes, while normally about 1 minute:

Waiting network configuration
Booting system without full network configuration

I found after googling that I should change /etc/networks/interface. I commented out everything there but the problem remains:

# auto lo
# iface lo inet loopback

# auto eth0
# iface eth0 inet dhcp
# address 192.168.0.2
# netmask 255.255.255.0
# gateway 192.168.0.1
# broadcast 192.168.0.255

How can you make the startup of Ubuntu 11.10 faster?

7
  • 5
    I do not understand this behavior. I do not want to change my interfaces file every time I go home. There are situations where I simply have no network OR I use wicd AFTER logging in. For me as a laptop owner that means waiting 2 senceless minutes. Twice every day. Besides: I configured all interfaces to be static. No way.
    – user32333
    Nov 4, 2011 at 22:12
  • 2
    @jrg It is not a code problem, it a desing problem. Can I file a bug about such a problem? Nov 4, 2011 at 22:53
  • 1
    Yes, you can. Just make a note that it is a design bug. :)
    – jrg
    Nov 5, 2011 at 0:33
  • @jrg What should be the status of the bug? - I selected xorg as the package, since needed to select something. I put the status to invalid to it, since it is not really about it. - It is about the design. Can people understand now that the status of the bug is not declined, but only for the package xorg? Nov 5, 2011 at 14:04
  • 2
    Use allow-hotplug instead of auto in the /etc/network/interfaces file, e.g., allow-hotplug enp0s25 then iface enp0s25 inet dhcp. This tells the server to not wait for the interface to light up with a dhcp answer. May 11, 2017 at 14:37

7 Answers 7

38

First off, this is a new behavior, documented in the 11.10 release notes, that I actually developed together with Scott Moser as an effort to make server boot more reliable.

Commenting out lo will mean you have no local network capability, which will break some programs when they try to use the network. It will also cause your system to never boot because it is so critical. So leave these two lines:

auto lo
iface lo inet loopback

The bits about eth0 meant that your machine was configured to wait for a dynamic address to be assigned to it before the network is considered UP. In pre-upstart versions of Ubuntu (8.10 and earlier), the system would have waited up to 60 seconds for this before continuing the boot. When upstart was added, this condition wasn't waited for anymore, because network interfaces that were not always expected to be plugged in are better managed by something like network-manager.

So, if you have a server, you probably want to wait for a dynamic address, otherwise the system will boot without all of its networks available (which it does if it takes more than 2 minutes to get an address). If you have a laptop that you don't always expect to be plugged in to eth0, then configure eth0 in network manager, and remove only those lines from /etc/network/interfaces, which should get rid of your boot delay.

Keep in mind, there's a known bug with VMware and dbus that also causes this message.

19
  • 1
    Masi, sounds like you need to ask a different question.
    – SpamapS
    Oct 19, 2011 at 0:24
  • 3
    Sorry, I am a newbie and I still don't understand why the system must wait around 2-4 extra minutes for IP. And the main question is "do I need this one. if I have a laptop and I always use wired and wireless network connections?"
    – itun
    Nov 2, 2011 at 10:14
  • 2
    @SpamapS and Scott Moser Please, change the thing back it was. You are hurting Desktop users. I understand your point that you save time in programming things, but please consider if 100 million users lose every day 2 minutes because of such a change. It is a huge cost! Please, branch Ubuntu Server completely different from Ubuntu Desktop. Nov 4, 2011 at 22:50
  • 5
    upvoting this for awareness BUT as a normal Desktop user I am VERY UNHAPPY that there is no fast boot option for this. Definitely perceived as a degradation from 10.04 to 12.04 upgrade. This problem is compounded by the fact that network manager is unable to directly control settings such as duplex and autoneg, resulting in the need to achieve that via /etc/network/interfaces
    – prusswan
    Feb 6, 2012 at 4:17
  • 2
    @SpamapS: While I'm glad you've pondered adding a no wait option, what really needs to be pondered is why Canonical so consistently fails at QA (user for the last 6 years, and I encounter showstopper regressions every release that should have prevented every release). That aside, the server and minimal CDs are widely used for desktop installation because the graphical installer often has issues and/or it's easier getting the minimal CD running on faulty media. It is definitely not safe to assume that because a user is using the alternate or minimal CD they must be installing a server. Sep 22, 2012 at 6:21
8

If you moved your OS from one machine to another it might be a good guess that udev created a configuration containing the mac address of your old network device which is different from the mac address of the network device from your new system.

Try removing the following file then reboot your system and see if this solves the problem:

/etc/udev/rules.d/70-persistent-net.rules
1
  • dachange thanks for getting me on the right track to solve a different issue ! i was getting a reset timeout on my wlan0 rlt8185 device when executing ip link set wlan0 up. Looked in the 70-persistent-net.rules file and commented out the line that added my wireless device, rebooted and now it's working!
    – user71562
    Jun 18, 2012 at 19:56
6

You're mixing 2 incompatible ways of assigning an address to a network interface.

iface eth0 inet dhcp

says "send out a DHCPDISCOVER packet to the physical local network, wait for a DHCPOFFER packet from a physically local DHCP server, and get the IP address from it (see http://www.rfc-editor.org/rfc/rfc1531.txt), while the rest of the lines assign values to the interface "manually". It's important that one gets the IP address assignment method right. If there is a DHCP server on your local network, you should use it. To do so, delete the "address", "netmask", "gateway", and "broadcast" lines.

If you're not connected to an "administered" local area network (and don't have a DHCP server), and want to assign the IP address parameters manually, change the first line to:

iface eth0 inet static

and keep the "address", "netmask", "gateway", and "broadcast" lines. See http://www.rfc-editor.org/rfc/rfc5735.txt for details about which IP addresses are available for use. The parameters you have look OK to me.

Read

man 5 interfaces

simply commenting everything out is not the best path to happiness.

10
  • Thank you for your answer! - Why does my Ubuntu work with empty /etc/network/interfaces? - How can you see that it is not working normally or efficiently? Sep 28, 2011 at 22:00
  • So the last line says: send out a static package to the physical local network, wait for a static package from a physical local network, and get the IP address from it. Sep 28, 2011 at 22:02
  • I have know [code]iface eth0 inet static address 192.168.0.2 netmask 255.255.255.0 gateway 192.168.0.1 broadcast 192.168.0.255[/code], in my /etc/network/interfaces and the problem still persists: loading a lot. Sep 28, 2011 at 22:09
  • No. The line with "static" says "Get the IP address from the following lines" no network packets are sent.
    – waltinator
    Sep 28, 2011 at 22:23
  • 1
    I think I understand. Because your mobile broadband isn't configured when the boot sequence does its ifconfig -a during startup, you get the timeout.
    – waltinator
    Oct 2, 2011 at 18:23
5

Seems like for me, the best solution for this problem was found at this linux site

Basically, still calling /etc/init/failsafe.conf , but commenting out the two sleep calls that caused the delay. I don't really see why they were added, since my network is configured fine without the need for a delay.

3

I use

eth0 inet dhcp

I don't see the point of waiting when there is no link (the cable is missing) There is no way dhcp can get the address if there is no media. This is a bug

1
  • 1
    i use the same code, but i still have to wait for network configuration every time. Jan 7, 2012 at 12:39
2

I recently just had this same problem. I tried to go in and comment out the sleep time in /etc/init/failsafe.conf file and ended up with the system just continuously trying to boot. I fixed this error by booting into safe mode and droping into a root shell and doing the following so that I had rw privleges:

mount -o remount,rw / 

I then issued the following command which brought up the file so that I could edit it back to it's orginal state:

sudo nano /etc/init/failsafe.conf 

Save and exit, then reboot the system.

2
  • Can you, please, say what you edited in the file failsafe.conf. Did you comment out the sleep calls? I tried that some years ago unsuccessfully. Jul 8, 2014 at 20:53
  • 2
    Absolutely. After you are in the failsafe.conf txt, you just look for the sleep calls. I put mine back to 5, 40, and 59 respectively. I however was unable to succesfully comment out the sleep calls as it caused the system to go into continuous boot.
    – clrn0979
    Jul 19, 2014 at 7:35
2

Use allow-hotplug instead of auto in the /etc/network/interfaces file, e.g., allow-hotplug enp0s25 then iface enp0s25 inet dhcp. This tells the server to not wait for the interface to light up with a dhcp answer. - - Dr. Zim's comment.

1
  • 1
    If i could up vote this answer twice i would! Thank you :) Nov 16, 2017 at 19:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .