Why do I keep losing connection to my Linode after 12 hours?

Hey there,

It looks like this may be caused by udev and systemd-networkd both trying to control the naming scheme of your network interfaces.

Can you try running echo 'GRUB_CMDLINE_LINUX="net.ifnames=0"' >>/etc/default/grub, and then rebooting your Linode? It looks like this should prevent udev from trying to apply the predictable network interface names which should hopefully resolve this for you.

Let us know if this helps!

Regards,
Ryan L.
Linode Support Staff

Done. The system booted fine.
I will wait for a day and check back (since that is the only way to see if it's working unfortunately.)

(Also, if anyone has the same problem and is reading this thread, make sure to run grub-mkconfig -o /boot/grub/grub.cfg afterwards. Though it might be an Arch only thing.)

So far about 18 hours have passed, and it's working!
Thank you @rl0nergan and the linode team very, very much.

Uh oh. It's happening again.
The MTR test gives me the "network unreachable" error.
After rebooting, I could confirm that the MTR works again, and networkctl output is different:

2 eth0 ether routable configured

Does this information help?

Hi @rl0nergan, could you help me?

Hey @Zaen,

I'm taking another look at into it now and trying to recreate it on my end. Hopefully I can find some more information for you that might be able to help. I'll follow up with you when I have some more info.

Ryan L.

This may be unrelated but….

On another VPS provider (not Linode) I lost my IPv4 connection every 12 hours.

The VPS was configured to use DHCP which gave a lifetime of - you guessed it - 12 hours.

The networkctl output resembled yours:

IDX LINK TYPE OPERATIONAL SETUP
1 lo loopback carrier unmanaged
2 eth0 ether routable failed

Because eth0 was 'failed', systemd didn't issue the renewal and the IP address lease expired.

In my case, it was due to a route conflict it was receiving from the DHCP server, and I had to tell netplan to ignore DHCP routes, which prevented the error when it first obtained the IP, and allowed systemd to see that all was OK with the adapter and it renewed the lease happily.

Linodes by default (I believe) don't use DHCP, so this likely doesn't apply, but I thought I'd share just in case.

Hey @Zaen,

I've been trying to recreate this on my end, but unfortunately with no success. Did you check out Andy's suggestion to see if that might help?

Ryan L.

I didn't notice the replies, sorry to both @andysh and @rl0nergan.
Is there a way to subscribe to this thread? (email)

I don't know what netplan is, but I am in fact using dhcpcd. I installed and enabled it when I installed Arch, and left it that way ever since. Is that the problem?

I checked the systemd log, and there was some normal 'deleting address' and 'carrier acquired' stuff going on, but as I have just rebooted my linode, the problem is not happening right now. I will post the dhcpcd logs after the problem occurs again.

I don't know what netplan is

Netplan is an Ubuntu thing; it allows defining networking in YAML files, and writes out the relevant config files for different backends.

I checked the systemd log, and there was some normal 'deleting address' and 'carrier acquired' stuff going on, but as I have just rebooted my linode, the problem is not happening right now. I will post the dhcpcd logs after the problem occurs again.

In my case, systemd didn't even try to renew the lease. It was a failure in the initial config of the adapter when the system was first booted. systemd was obtaining a route from the DHCP server that it believed was invalid.

Because of the error, systemd believed the adapter wasn't fully configured so wasn't watching for when the IP expired and didn't request a new IP from DHCP when it should have done.

In my experience with Linode, you are much better off configuring addresses statically; they won't change.

https://www.linode.com/docs/networking/linux-static-ip-configuration/

Hey @Zaen,

No worries! That's actually a great idea, I can see how that would be helpful. I've passed that feedback along to our developers to see if that's something we could implement in the future.

Ryan L.
Linode Support Staff

@andysh, @rl0nergan So sorry for going radio silent once again…
Weirdly (or not), the problem disappeared! Or rather, the network fails sometimes, and after that it just recovers. I'm guessing that's the case from looking at my Seafile (a cloud server thingy) client, like it says last updated 4 days ago, but when I refresh it, it starts working again - so the problem must have happened 4 days ago, and seafile tried reaching the server a bunch of times before giving up.

@andysh I have bookmarked the link, and I will use it in the future when I'm fixing this problem once and for all.
Because of the whole Covid thing I'm working outside less, so my motive to fix this is very low. I know I asked this question, but… I'm lazy :( I really appreciate both your help, please don't get me wrong.

I also looked everywhere to enable email notification for forum post, but I can't find it.