In the previous post, I mentioned that I kept an eye on the network traffic on my LTE connection on my laptop. I’ve been bitten hard (It updated IntelliJ IDEA, several hundred megabytes) by snap’s autorefresh feature, which have seemed hard/impossible to disable. There’s been config options for only allowing updates at certain times etc., but the feature I’ve been crying for, to be able to disable it when on a metered connection, has been missing in action. Until now!

$ snap set system refresh.metered=hold

Easy as that!

(Found it here - otherwise not much info about this around the web) UPDATE 2020-02-11: Link is dead, but this is in the official documentation here

Seems quite recent, introduced in snapd 2.33, which was released on 2018-06-25.

So how does it work? How do I confirm that it’s actually working? No feedback when running the command, nothing in syslog.

Hmm.

Only one thing to do; use the source, Luke!

From the source, I can see that it reads the DBus property org.freedesktop.NetworkManager.Metered for a value, and decides what to do based on that.

const ( 
        // https://developer.gnome.org/NetworkManager/stable/nm-dbus-types.html#NMMetered
        NetworkManagerMeteredUnknown  = 0
        NetworkManagerMeteredYes      = 1
        NetworkManagerMeteredNo       = 2
        NetworkManagerMeteredGuessYes = 3
        NetworkManagerMeteredGuessNo  = 4
)

If it’s a NetworkManagerMeteredYes or NetworkManagerMeteredGuessYes, it will disable autorefresh.

I then read the DBus property with the tool D-Feet, and confirmed that the value returned was ‘3’ when on LTE, and ‘4’ on Wi-fi. Wow, this actually seems to work!

While I would prefer more information about this (regular users will NEVER find it, and it is default off), and more logging, I’ll take this for now!

Sometimes I get curious about how much data a network interface has transferred. Actually, I get curious quite frequently after I got an LTE modem for my laptop.

Usually, I just use ifconfig:

$ ifconfig wlp3s0
wlp3s0    Link encap:Ethernet  HWaddr 00:15:00:7e:af:bd  
          inet addr:192.168.0.165  Bcast:192.168.0.255  Mask:255.255.255.0
          inet6 addr: fe80::4c15:ef76:132e:768/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:20922893 errors:0 dropped:0 overruns:0 frame:0
          TX packets:20916831 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000 
          RX bytes:15370202859 (15.3 GB)  TX bytes:13050802316 (13.0 GB)

With Ubuntu 18.04, however, there is no ifconfig. Yes, I know I can install it, but I wanted to learn how I should be doing it.

I came up with this, using the ‘ip’ command:

$ ip -s -h addr show wlp3s0
3: wlp3s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    link/ether 00:15:00:7e:af:bd brd ff:ff:ff:ff:ff:ff
    inet 192.168.0.165/24 brd 192.168.0.255 scope global dynamic wlp3s0
       valid_lft 5439sec preferred_lft 5439sec
    inet6 fe80::4c15:ef76:132e:768/64 scope link 
       valid_lft forever preferred_lft forever
    RX: bytes  packets  errors  dropped overrun mcast   
    15.4G      20.9M    0       0       0       0       
    TX: bytes  packets  errors  dropped carrier collsns 
    13.1G      20.9M    0       0       0       0

Meh. This is gonna take some time to get used to.

After upgrading LXD from 2.0.x to 3.2, with everything at first looking hunky dory, I discovered after a few days that I couldn’t create new containers any more!

$ lxc launch ubuntu:bionic testcontainer
Error: Failed container creation: No root device could be found

WTF!

On another host which received the same upgrade, however, everything still worked fine. After checking if there were any differences in the configuration, I found something missing in the default profile, in the devices section.

On the host where everything worked:

devices:
  eth0:
    name: eth0
    nictype: bridged
    parent: br0
    type: nic
  root:
    path: /
    pool: default
    type: disk

On the one which DIDN’T work:

devices:
  eth0:
    name: eth0
    nictype: bridged
    parent: br0
    type: nic

AHA! The entire ‘root’ section is gone! Why didn’t it get added in the first place? I don’t know.

Adding it was easy though, just did “lxc profile edit default” and pasted it in, and everything works fine.

Weird stuff.

I’ve done this a few times now, and when you know how to do it, it’s a breeze.

Posting the “recipe” here, for my self and anyone who might need it.

Should be pretty self explanatory for those with a little experience with ZFS.

Obviously, after starting LXD after migrating, you should test if everything is OK before destroying the old dataset.

(you will need the packages sqlite3 and mbuffer, mbuffer can be omitted at the cost of some speed)

When installing Ubuntu with root on ZFS (following this great guide), I soon discovered that something was amiss with the networking. Wi-fi worked fine, but my mobile broadband failed to connect, and NetworkManager didn’t respond when I plugged in an ethernet cable.

After literally HOURS of troubleshooting (beginning in the wrong end with the mobile broadband), I stumbled across this bug report. Sounds familiar, I installed the ZFS root with a chroot (debootstrap). Thankfully, the comments have the solution.

Create an empty file, and restart stuff.

$ sudo touch /etc/NetworkManager/conf.d/10-globally-managed-devices.conf
$ sudo systemctl restart NetworkManager ModemManager

Voilà! Problem solved.

After about seven months of waiting, I can soon enjoy tomatoes from the plant I planted from seed in February!

LXD containers are brilliant, but lacks an easy way to forward ports from the containers to the host. One can use iptables manually, of course, but I really missed something easy like Docker. To try and remedy this, I have conjured up a little bash script. With this script, you can add, delete and list port forwarding rules.

It’s a bit rough around the edges, but maybe I’ll tidy it up a bit some day.

Read on for the script and examples.