K12 here.. don't hate; limited budget, etc..
I have a fleet of bare metal machines with an on board e1000e Intel
card. (copper 1Gb)
Which I have been using as a vlan enabled device:
ip link | grep -c eth0
14
With the ifupdown-ng-iproute2 syntax:
auto lo
iface lo inet loopback
auto eth0
auto eth0.42
address 10.120.48.247/24
gateway 10.120.48.1
auto eth0.10
address 10.20.48.247/20
auto eth0.49
address 10.120.49.247/24
auto eth0.50
address 10.120.50.247/24
auto eth0.172
address 172.16.48.247/24
etc..
(thank you everyone who pays taxes in my US State.. )
I was able to get an order of Intel X710-DA2's (uses i40e dual sfp+)
I had no way in the bios to disable the onboard nic.. so the new i40e
card came up as eth1 and eth2 (as expected)
But all my ansible scripts and configs are for eth0 (copper) and this
card is sfp+ (dac in my case and only eth0 - cage A should it matter)
So I thought I would just blacklist the e1000e module (thereby no
driver for the onboard card) and my eth1 & eth2 would be promoted to
eth0 & eth1 (which of course it did.. )
Box comes up (hmm stalls? pauses?; slower) but sub interfaces load and
we're passing traffic; great..
Obviously don't reboot much, but with all the 'AI kernel hacking'
reboots are more frequent.. (thank you for any and all security team
work.. )
In trying to figure out the boot slowness.. it seems the sub
interfaces incur a 1 ~ 1.5 second penalty or stall/wait when loading..
with e1000e blacklisted
cat /etc/modprobe.d/e1000e.conf
blacklist e1000e
# install e1000e /bin/false
(the install didn't make any difference either)
Tried rc_parallel="YES" in /etc/rc.conf ( no difference )
I am open to "blacklisting the module was a bad idea; do $this instead"..
dmesg | grep eth0
[ 3.711847] i40e 0000:01:00.0 eth0: NIC Link is Up, 10 Gbps Full
Duplex, Flow Control: None
[ 6.311248] 8021q: adding VLAN 0 to HW filter on device eth0
[ 27.056801] i40e 0000:01:00.0 eth0: Changing Tx descriptor count
from 512 to 4096.
[ 27.056907] i40e 0000:01:00.0 eth0: Changing Rx descriptor count
from 512 to 4096
[ 29.743908] eth0: renamed from veth8e9e2f0
# cat /proc/cmdline
BOOT_IMAGE=/boot/vmlinuz-lts
root=UUID=d741f826-8319-47bd-810b-11b8b7df633d ro
modules=sd-mod,usb-storage,xfs quiet rootfstype=xfs mitigations=off
skew_tick=1 modprobe.blacklist=e1000e,e1000
# EOF
Can live with if need be.
Thank you in advance.
Please let me know if anything else would be helpful.
https://wiki.alpinelinux.org/wiki/Configure_Networking#Advanced_network_configuration
Setting fixed interface names can prevent unexpected network configuration
issues and the above section has few suggestions on it.
Generally loading/blacklisting modules shouldn't cause slowness, as it
might be more due to ip configuration.
On Sun, 31 May, 2026, 8:18 AM B. Cook, <bcook@poughkeepsieschools.org>
wrote:
> K12 here.. don't hate; limited budget, etc..>> I have a fleet of bare metal machines with an on board e1000e Intel> card. (copper 1Gb)>> Which I have been using as a vlan enabled device:>> ip link | grep -c eth0> 14>> With the ifupdown-ng-iproute2 syntax:>> auto lo> iface lo inet loopback>> auto eth0>> auto eth0.42> address 10.120.48.247/24> gateway 10.120.48.1>> auto eth0.10> address 10.20.48.247/20>> auto eth0.49> address 10.120.49.247/24>> auto eth0.50> address 10.120.50.247/24>> auto eth0.172> address 172.16.48.247/24>> etc..>> (thank you everyone who pays taxes in my US State.. )>> I was able to get an order of Intel X710-DA2's (uses i40e dual sfp+)>> I had no way in the bios to disable the onboard nic.. so the new i40e> card came up as eth1 and eth2 (as expected)>> But all my ansible scripts and configs are for eth0 (copper) and this> card is sfp+ (dac in my case and only eth0 - cage A should it matter)>> So I thought I would just blacklist the e1000e module (thereby no> driver for the onboard card) and my eth1 & eth2 would be promoted to> eth0 & eth1 (which of course it did.. )>> Box comes up (hmm stalls? pauses?; slower) but sub interfaces load and> we're passing traffic; great..>> Obviously don't reboot much, but with all the 'AI kernel hacking'> reboots are more frequent.. (thank you for any and all security team> work.. )>> In trying to figure out the boot slowness.. it seems the sub> interfaces incur a 1 ~ 1.5 second penalty or stall/wait when loading..> with e1000e blacklisted>> cat /etc/modprobe.d/e1000e.conf> blacklist e1000e> # install e1000e /bin/false>> (the install didn't make any difference either)>> Tried rc_parallel="YES" in /etc/rc.conf ( no difference )>> I am open to "blacklisting the module was a bad idea; do $this instead"..>> dmesg | grep eth0> [ 3.711847] i40e 0000:01:00.0 eth0: NIC Link is Up, 10 Gbps Full> Duplex, Flow Control: None> [ 6.311248] 8021q: adding VLAN 0 to HW filter on device eth0> [ 27.056801] i40e 0000:01:00.0 eth0: Changing Tx descriptor count> from 512 to 4096.> [ 27.056907] i40e 0000:01:00.0 eth0: Changing Rx descriptor count> from 512 to 4096> [ 29.743908] eth0: renamed from veth8e9e2f0>> # cat /proc/cmdline> BOOT_IMAGE=/boot/vmlinuz-lts> root=UUID=d741f826-8319-47bd-810b-11b8b7df633d ro> modules=sd-mod,usb-storage,xfs quiet rootfstype=xfs mitigations=off> skew_tick=1 modprobe.blacklist=e1000e,e1000>> # EOF>> Can live with if need be.>> Thank you in advance.>> Please let me know if anything else would be helpful.>
I’m sorry if what I wrote was not clear.
It seems interface vs device would be better understood if I clarified..
There are no network delays on boot or in practice with e1000e device
loaded and all the sub interfaces utilizing the onboard device.
blacklisting e1000e allows the i40e driver to put the x710 as eth0 and
eth1; is the addition of eth1 and being undefined? There was no delay with
eth1 and eth2 but not defined.. (need to confirm)
I am not renaming interface (as you are saying) rather disabling unwanted
*devices*.
Alpine “naturally” creates eth0 as primary interface from the first found
device.
If I could disable this device in the bios, and start the machine with no
network adapter I would expect networking to complain about the inability
to find eth0; this would make sense; no device - no module loaded - no eth0
The kernel is finding eth0 as administratively desired established by i40e
module and device and not e1000e (again inability to disable on board
device)
But now this comes with a delay or stall.. not administratively desired or
designed..
Does that further help with the problem description?
Please let me know
On Sat, May 30, 2026 at 11:13 PM Prabu Anand Kalivaradhan <
kprabuanand@gmail.com> wrote:
>> https://wiki.alpinelinux.org/wiki/Configure_Networking#Advanced_network_configuration>> Setting fixed interface names can prevent unexpected network configuration> issues and the above section has few suggestions on it.>> Generally loading/blacklisting modules shouldn't cause slowness, as it> might be more due to ip configuration.>> On Sun, 31 May, 2026, 8:18 AM B. Cook, <bcook@poughkeepsieschools.org>> wrote:>>> K12 here.. don't hate; limited budget, etc..>>>> I have a fleet of bare metal machines with an on board e1000e Intel>> card. (copper 1Gb)>>>> Which I have been using as a vlan enabled device:>>>> ip link | grep -c eth0>> 14>>>> With the ifupdown-ng-iproute2 syntax:>>>> auto lo>> iface lo inet loopback>>>> auto eth0>>>> auto eth0.42>> address 10.120.48.247/24>> gateway 10.120.48.1>>>> auto eth0.10>> address 10.20.48.247/20>>>> auto eth0.49>> address 10.120.49.247/24>>>> auto eth0.50>> address 10.120.50.247/24>>>> auto eth0.172>> address 172.16.48.247/24>>>> etc..>>>> (thank you everyone who pays taxes in my US State.. )>>>> I was able to get an order of Intel X710-DA2's (uses i40e dual sfp+)>>>> I had no way in the bios to disable the onboard nic.. so the new i40e>> card came up as eth1 and eth2 (as expected)>>>> But all my ansible scripts and configs are for eth0 (copper) and this>> card is sfp+ (dac in my case and only eth0 - cage A should it matter)>>>> So I thought I would just blacklist the e1000e module (thereby no>> driver for the onboard card) and my eth1 & eth2 would be promoted to>> eth0 & eth1 (which of course it did.. )>>>> Box comes up (hmm stalls? pauses?; slower) but sub interfaces load and>> we're passing traffic; great..>>>> Obviously don't reboot much, but with all the 'AI kernel hacking'>> reboots are more frequent.. (thank you for any and all security team>> work.. )>>>> In trying to figure out the boot slowness.. it seems the sub>> interfaces incur a 1 ~ 1.5 second penalty or stall/wait when loading..>> with e1000e blacklisted>>>> cat /etc/modprobe.d/e1000e.conf>> blacklist e1000e>> # install e1000e /bin/false>>>> (the install didn't make any difference either)>>>> Tried rc_parallel="YES" in /etc/rc.conf ( no difference )>>>> I am open to "blacklisting the module was a bad idea; do $this instead"..>>>> dmesg | grep eth0>> [ 3.711847] i40e 0000:01:00.0 eth0: NIC Link is Up, 10 Gbps Full>> Duplex, Flow Control: None>> [ 6.311248] 8021q: adding VLAN 0 to HW filter on device eth0>> [ 27.056801] i40e 0000:01:00.0 eth0: Changing Tx descriptor count>> from 512 to 4096.>> [ 27.056907] i40e 0000:01:00.0 eth0: Changing Rx descriptor count>> from 512 to 4096>> [ 29.743908] eth0: renamed from veth8e9e2f0>>>> # cat /proc/cmdline>> BOOT_IMAGE=/boot/vmlinuz-lts>> root=UUID=d741f826-8319-47bd-810b-11b8b7df633d ro>> modules=sd-mod,usb-storage,xfs quiet rootfstype=xfs mitigations=off>> skew_tick=1 modprobe.blacklist=e1000e,e1000>>>> # EOF>>>> Can live with if need be.>>>> Thank you in advance.>>>> Please let me know if anything else would be helpful.>>>