I seem to have lost the ethernet connection to my desktop. There is nothing to indicate a problem; Fedora tells me it’s connected, but I can’t ping anything. Not sure how to troubleshoot. I’ve tried switching cables, but it doesn’t help. But I know that there’s no problem with the router, because wireless is up (I’m typing this from my notebook).
(Monday-morning update)
Still haven’t solved it, but on a flight to Denver for suborbital researchers conference, so won’t get to it until Thursday.
[Thursday-morning update]
It’s clearly a weird DNS problem. I can ping Google’s IP, but I cannot ping google.com. I can ping 8.8.8.8, but when I manually set that as DNS, it still doesn’t work.
I tried plugging directly into the router instead of the Orbi, with same results.
I have not changed anything in the configuration; it had been working fine for many months as it was set up, with the router feeding the main Orbi via ethernet, and the Orbi providing both ethernet and wifi to my desktop and the house. But for some reason, I woke up almost a week ago, and the desktop had no Internet.
[Bumped]
[Afternoon update]
Well, this is fun. It’s clearly a software issue.
I just pulled the SSD off my laptop and booted the desktop from it, and it now works. I’m suspecting that it has something to do with expressvpn. Since I installed it, I don’t have Internet until I connect it. But on the machine that has problems, I couldn’t connect, so I removed it, but it didn’t solve the problem. There must be some setting that doesn’t allow the machine to do DNS without expressvpn, which somehow went haywire last week.
I’m wondering if the best solution would be to just install a clean version of Fedora 31 (I’m currently using 30) on either the old SSD, or a new one.
[Friday update]
OK, it turned out that ExpressVPN was locking out the Internet when it wasn’t connected (for my “safety”), which meant that if something went wrong with the connection, it made it impossible to reconnect to ExpressVPN, or the Internet. After spending several hours this morning on chat with someone at ExpressVPN, and wasting my time going down various rabbit holes, I finally found someone who told me to just blow away /etc/resolv.conf, and reconnect, which gave me Internet again.
I did manage to convey that they had some serious bad practices that resulted in this frustration, and gave them some recommendations.
- Let the icon tell me whether I have a good connection. Ever since I installed the expressvpn package, it always has a question mark on it, whether it is connected or not.
- Have an autoconnect that actually works on login, rather than having to go into shell each time and manually connect.
- Provide an ability to override the lock (with a warning) rather than simply lock it when ExpressVPN isn’t connected.
They said they’d try to incorporate into the next version.
Reboot the desktop? Or if you prefer to tinker try toggling the interface up and down. Always try the stupid stuff first. Do you have anything else that is wired that is working?
Just to eliminate the wired part, plug your notebook into a wired port and see if you can connect. I’m guessing you can.
I’ve toggled on and off, I’ve tried different cables, and I’ve rebooted. No joy. My notebook has no ethernet port. I may have to borrow her work laptop to test it the cable.
Well there is always power cycling the router, just in case it’s a firmware issue. If you have a separate cable modem, the procedure I follow is power off router first then cable modem. Wait 30 seconds then power on cable modem, wait about 60 seconds and then power up the router. If you have an integrated one (like the kind the cable cos provide) then there is only one device to power cycle. You may then have to power cycle / reboot the downstream devices. Unless you are running a server farm, it’s not that big a deal. About once a year my cable co does something that requires this action. Usually to get soft data corrected in the modem or router.
I know my Orbi mesh is getting packets from the cable router (or I wouldn’t have wireless on my notebook), so it’s unlikely that rebooting it will help. The only way to test would be to move the cable end from the Orbi to my desktop, but that would momentarily kill the wifi, and she’s using her computer right now.
It’s unusual to lose the wired part of the router but I suppose it’s possible. Yeah don’t kill the Internet when the better half is on-line. I learned that lesson the hard way.
I still say you should reboot the router as a last resort. Also usually I take the opportunity to make sure my router’s firmware is up-to-date.
Just to be clear, can you ping the router? Are you using DHCP or a manual IP? What about DDNS?
I’d try setting up the desktop with a manual IP in the range allowed by the router. Then try pinging the router’s IP address.
Can you see LED activity on the desktop ethernet? If no joy, you might have dead ethernet either at the desktop or the router.
OK, so I’ve rebooted router, and mesh router and its satellite. The desktop is plugged into the mesh router. I can see the network (i.e., I can ping the mesh router, and I can ping my notebook). But I have no Internet, though clearly the mesh router does, or I wouldn’t have Internet from my notebook. I guess I should be happy that my network interface is working, but it doesn’t solve my problem.
Is the desktop computer the only thing using Ethernet on your network?
If I were troubleshooting a similar problem, I would see whether another wired device can reach the internet (an excellent reason to keep at least two computers with Ethernet ports, even if one could run only Windows 8). If not, I would suspect a firmware glitch in the router. But if the only computer unable to reach the internet is your desktop, it would have to be in the computer.
So, ethernet is fine, IP is mostly fine. Progress.
Check the routing table (“route” as root should do it).
Can you ping the google public DNS by IP (8.8.8.8)?
If not, where does traceroute to it fail (“traceroute 8.8.8.8”).
On a Windows box I might try disabling the Ethernet in the hardware manage, wiping it out as thoroughly as possible, and then let Windows rediscover it. I’m not sure how Linux handles new devices, though. There was a generation or so of PC’s where just moving an Ethernet card to another slot would let the OS go through a “found new hardware” setup routine.
And of course there’s the time-tested tactic of Googling whatever error message you’re getting, on the theory that you’re certainly not the first person to have the problem. But you’ve probably already done that.
Did you allow Fedora to update or patch within the last day?
This is admittedly a longshot due to a different OS, but, the reason I ask is that I had a client a few years ago, whose OS suddenly become unable to see the onboard ethernet. It was because they’d allowed their OS to update whatever it felt like, and in this case, it made Windows unable to see the onboard ethernet. There was no option to backtrack, so the fix ended up being adding an ethernet card (it was that, or reinstall the whole OS).
For this and a lot of other reasons, I always advise against automatic updates. Far better to review and see what’s actually needed; if it ain’t broke, don’t try to fix it.
I hope all goes well, good luck.
Try to ping 216.58.194.110 (Google). If it pings the raw address it is likely a DNS problem. Also look at the leds on the ports, usually green for connection and flashing amber for traffic.
Running ifconfig will give you what your computer thinks is the gateway and DNS server. you can ping them to see if you can connect. You can change the DNS server to 8.8.8.8 (again Google) and see if that helps.
This is a list of other DNS servers:
https://www.lifewire.com/free-and-public-dns-servers-2626062
Two routers one network. Reminds me of the cuneiform symbol for trouble which decomposes into two women in the same house (under the same roof). Make sure the Mesh router and the cable router are not both acting as DHCP servers. There should only be one active on a given network. You can test this by assigning a static up to your desktop using a subnet number that is far from what is being doled out by DHCP but remains routable on your LAN. So if your wireless are getting assigned addresses like 192.168.1.100 .101 .102 try a static up address like 192.168.1.150 on the desktop and see if it starts to work. What you really should be doing is configuring the Orbi to act like a bridge not a router. In other words there should only be one gateway and that is your cable router. OR run a long Ethernet cable to your cable router to see if you can get the desktop machine on line that way. Did this work before or is the Orbi a new addition? Rand you sure have a knack for getting crossways with electric tech. Maybe like Kelly Johnson you’d have better luck with hydraulics.
Netgear recommends running the Orbi in access point mode. Do a network search using this phrase
“configuring an orbi as a bridge” for pointers. Maybe you are doing this already? Let’s hope.
Here’s a link… You are not alone…
https://kb.netgear.com/31218/How-do-I-configure-my-Orbi-router-to-act-as-an-access-point
If your cable co router has WiFi it is also recommended you turn it off so the two do not interfere with each other.
So putting together the facts:
Your machine has a IP address.
You can ping the internal router ip address.
Suggest trying to Ping a IP address like MCS suggested to verify that it not a DNS issue
May want to try “traceroute 1.1.1.1” to see how far down the line you go to see what router is stopping the packet.
May want to try “route” to verify the machine default gateway is the router
Have you logged into the router/ and or Mesh router admin to verify nothing is off in the settings and that you can.
Other thought is have you tried a cross over/non cross over network cable , maybe the automatic cross over on the network card or mesh router failed (Though you shouldn’t have been able to ping the router)
I you can ping the DNS server, but not get DNS queries to/from it then it is a firewall or DNS issue. Since you say that you just uninstalled a VPN, almost certainly a DNS issue.
Look at /etc/resolve.conf – you probably have something in there that is rerouting DNS queries. See:
https://github.com/hkuchampudi/Windscribe/issues/3
OK, but I uninstalled the VPN after the problem started, not before.
Could you somehow have a bad subnet mask (255.255.255.0 became something else?) Could an entry be screwed up in your /etc/networks, protocols, services, or hosts files? The time/date stamp should show if they were changed at about the same time the problem appeared.
Since you said the problem occurred after uninstalling Expressvpn, try this:
https://unix.stackexchange.com/questions/339334/after-expressvpn-uninstall-no-connection-to-working-dns-service-anymore-after-u
Out of idle curiosity, why do you use Fedora as opposed to another distro?
Inertia.
Well, this is fun.
For some definition of the word “fun”.
Now that we know all that you describe, I agree with Rick C this looks good. Try what’s recommended there (fixing up resolv.conf), then toggle the interface and see if it comes back. It shouldn’t matter what port it’s plugged into given your history, but sometimes its helpful to know it works when working backwards, i.e. cable modem wired connection first then Orbi. That shouldn’t matter, but when I’m troubleshooting I like to be pedantic to know what IS working first.
The hardware and network are pretty clearly fine, and you probably even have fine internet, it sounds like maybe DNS is all that’s broken.
Hard to say more without knowing how you were setting it. Maybe the VPN left some crazy firewall nonsense in place, you could have a look at “iptables -L” to see if there’s anything suspicious compared to your other Linux boxes…