Bug#931130: Document that local configuration for dummy and bonding modules are getting overwritten by systemd

Baptiste BEAUPLAT lyknode at cilg.org
Fri Jun 28 13:46:15 BST 2019


On 6/28/19 2:39 PM, Michael Biebl wrote:
> Am 28.06.19 um 12:58 schrieb Justin B Rye:
>> Michael Biebl wrote:
>>> Afaiu, the kernel default is to create one dummy device by default when
>>> the module is loaded.
> 
> .. or bond device, for that matter.
> 
>>> I assume there might be cases where users rely on that default behaviour
>>> without having explicitly configured anything.
>>
>> Do we have any idea what their use case would be?  Apparently a
>> different one from Baptiste's, if it's true that he needed to
>> explicitly configure "numdummies=1"...
> 
> Dunno why Baptiste had to do that.

I ended up creating the file to explicitly configure numdummies=1 when I
saw that my interfaces disappeared after the upgrade to buster.

This "customization" is only necessary when multiple interfaces are
required to show on boot.

> I just verified that by removing /lib/modprobe.d/systemd.conf, I got the
> following when running "modprobe dummy" and "modprobe bonding":
> 
> 6: bond0: <BROADCAST,MULTICAST,MASTER> mtu 1500 qdisc noop state DOWN
> group default qlen 1000
>     link/ether 82:e5:b3:41:67:17 brd ff:ff:ff:ff:ff:ff
> 7: dummy0: <BROADCAST,NOARP> mtu 1500 qdisc noop state DOWN group
> default qlen 1000
>     link/ether 5e:eb:30:64:69:ef brd ff:ff:ff:ff:ff:ff
> 
> With the modprobe.d config file installed, there obviously was neither a
> dummy0 nor bond0 device.
> 
> Fwiw, I know too little about those dummy and bonding devices and how
> they are usually used.
> E.g. network-manager and networkd seem to create them on the fly as
> needed. So I'm not sure, which users will actually be affected by that
> change.

All users not using networkd and network-manager. Usually servers, just
with ifup/ifdown.

> Would be good to know how Baptiste is setting up those devices. If he is
> doing it manually via some scripting of low level tool or uses a higher
> level network management tool
> Fwiw, with "ip link add dummy0 type dummy" (or "ip link add bond0 type
> bond") I was e.g. able to create such a device manually as well.
> I wonder whether such an approach isn't better then statically setting
> the number of devices via a kernel module option.

My original setup was:

echo "dummy" > /etc/modules

cat << EOF >> /etc/network/interfaces
auto dummy0
iface dummy0 inet static
    address 192.168.64.1
    netmask 24
EOF

The interface would pop up configured at boot time (by ifup). Then, I
had services binding on 192.168.64.1.

I hope this info helps.

Best,

-- 
Baptiste BEAUPLAT - lyknode

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: OpenPGP digital signature
URL: <http://alioth-lists.debian.net/pipermail/pkg-systemd-maintainers/attachments/20190628/e483cb55/attachment.sig>


More information about the Pkg-systemd-maintainers mailing list