3

Problem: 5 minute delay on startup

Device: UP board (small x86 board), running 18.04, with a Panda PAU05 wireless USB adapter, connected via HDMI to a monitor

Network configuration: using netplan and systemd-networkd. NetworkManager is not installed.

When the Ethernet is connected, the device boots up quickly. I can unplug the Ethernet and rely on wifi. SSH works, etc.

When the Ethernet is not connected, there's a 5 minute delay. The console says A start job is running for Raise network interfaces (Xmin Ys / 5min 2s) and slowly counts up to 5min 2s. I can ping the device, but ssh is refused: ssh: connect to host rodeobot.local port 22: Connection refused. After 5 minutes the boot completes fine.

I tinkered with my netplan to make both ethernet and wifi interfaces optional, but this doesn't help. I also tried removing the ethernet interface, but that didn't help. Here's the /etc/netplan/config.yaml I'm running now:

network:
  version: 2
  renderer: networkd
  ethernets:
    enp1s0:
      optional: true  # Don't wait for Ethernet
      dhcp4: true
  wifis:
    wlx9cefd5fcb328:
      optional: true  # Don't wait for Wifi
      dhcp4: true
      access-points:
        "xxx":
          password: "yyy"

Edit: I see several similar questions, but I don't see an answer that uses netplan. Is this a bug / missing feature in netplan?

Edit: fixed indentation and commenting in config.yaml. (I copied and pasted, then edited the ssid and pw.)

Edit: here's the result of sudo netplan --debug generate:

** (generate:965): DEBUG: 18:35:38.361: Processing input file /etc/netplan/config.yaml..
** (generate:965): DEBUG: 18:35:38.362: starting new processing pass
** (generate:965): DEBUG: 18:35:38.362: wlx9cefd5fcb328: adding wifi AP 'Beckett'
** (generate:965): DEBUG: 18:35:38.362: wlx9cefd5fcb328: setting default backend to 1
** (generate:965): DEBUG: 18:35:38.362: Configuration is valid
** (generate:965): DEBUG: 18:35:38.362: enp1s0: setting default backend to 1
** (generate:965): DEBUG: 18:35:38.362: Configuration is valid
** (generate:965): DEBUG: 18:35:38.363: Generating output files..
** (generate:965): DEBUG: 18:35:38.363: NetworkManager: definition enp1s0 is not for us (backend 1)
** (generate:965): DEBUG: 18:35:38.363: wlx9cefd5fcb328: Creating wpa_supplicant configuration file run/netplan/wpa-wlx9cefd5fcb328.conf
** (generate:965): DEBUG: 18:35:38.363: Creating wpa_supplicant service enablement link /run/systemd/system/systemd-networkd.service.wants/netplan-wpa@wlx9cefd5fcb328.service
** (generate:965): DEBUG: 18:35:38.364: NetworkManager: definition wlx9cefd5fcb328 is not for us (backend 1)

Thanks.

Clyde McQueen
  • 73

3 Answers3

1
network:
  version: 2
  renderer: networkd
  ethernets:
    enp1s0:
      optional: true
      dhcp4: true
  wifis:
    wlx9cefd5fcb328:
      optional: true
      dhcp4: true
      access-points:
        "xxx":
          password: "yyy"

sudo netplan generate # generate config files

sudo netplan apply # apply new configuration

reboot # verify proper operation

heynnema
  • 70,711
  • Still no luck. I added the results of sudo netplan --debug generate in case that helps. – Clyde McQueen Jun 03 '19 at 01:38
  • @ClydeMcQueen Temporarily comment out all of the wifis stanza, leave just the ethernets, netplan generate/apply, and retest. If that doesn't work, temporarily remove dhcp4, set a static address and gateway, and retest. Report back. – heynnema Jun 03 '19 at 02:54
  • Thanks for all of your help. My solution was to edit /etc/network/interfaces by hand, replacing auto enp1s0 with allow-hotplug enp1s0. I am surprised that sudo netplan apply didn't edit that file. I'm somewhat worried that my networking install is a bit broken. But now my boot times are quick w/ or w/o Ethernet plugged in. – Clyde McQueen Jun 04 '19 at 03:16
0

It turns out that enp1s0 wasn't managed by systemd-networkd, and therefore netplan had no effect. I discovered this by looking at the logs:

journalctl | grep enp1s0 | grep "managed by us"

I found a number of entries that looked like this:

systemd-networkd: enp1s0: Link is not managed by us

The fix was to remove the ifupdown package. For good measure (and to make sure I didn't forget!) I moved the /etc/network directory:

sudo apt remove ifupdown
sudo mv /etc/network /etc/network_old_ifupdown
sudo reboot now

While looking at the logs I also fixed a problem with resolv.conf -- I had a bad symlink. There's a good explanation here: New alert keeps showing up: Server returned error NXDOMAIN, mitigating potential DNS violation DVE-2018-0001

Thanks to folks that helped!

Clyde McQueen
  • 73
  • This is not the right fix for this problem. – heynnema Jun 08 '19 at 19:09
  • I'm not sure what you mean. It fixed it for me. – Clyde McQueen Jun 10 '19 at 00:50
  • That's because you were mixing Network-Manager and netplan commands. Your specific problem was with /etc/network/interfaces, and by renaming /etc/network, you just got that file out of the way. You should have just left /etc/network/interfaces as you found it originally, and not uninstalled portions of the OS to "fix" a different part of the system. You've handicapped your OS. – heynnema Jun 10 '19 at 01:06
  • Yes, this device had a tortured history: a clean 16.04 desktop install, then I attempted to remove all vestiges of the GUI when it became a headless device. I should have wiped it and installed 16.04 server. In fact, there's still time to wipe it an install 18.04 server, which might fix a few other things. Thanks again. – Clyde McQueen Jun 10 '19 at 19:34
0

Setting sudo nano /etc/network/interfaces

#from
auto lo br0
iface lo inet loopback

# to 
auto lo
allow-hotplug br0
iface lo inet loopback

helped. The slow startup occurred under Ubuntu 18.04 after following this tutorial to install KVM.

hb0
  • 221