2 of 3 PCs on same switch ... [SOLVED]

[dufeu]

edit: Sun Nov 26 12:55:19 EST 2017
TL:DR version: Verizon FiOS {now Frontier} provided router assigned multiple IP addresses to both PCs. Could not clear multiple assigned IP addresses from router's admin. Could not clear addresses using a hard reboot. Resolved by resetting router to factory defaults. I'm writing up a more complete post now and will link to it from here when it is complete.
end edit

I have 3 PCs and a printer on the same switch. Everything was working fine up until the last upgrade{from more than 4 weeks ago}. I didn't immediately notice the issue because I've only updated the kernel on my main {working}, not @world.

I now have two of the three PC unable to ping the router or the internet. The 3rd PC is still behaving perfectly fine.

The two problem PCs cannot ping the router. The router cannot ping the 2 problem PCs.

All the devices on the switch can see each other. The working PC and the printer are freely available to the rest of my local network.

Each of the same switch PCs can freely access the other PCs NFS and SAMBA shares.

In order to simplify diagnostics, I replaced the original configured bridging I had on one of the PCs and set it up with

[SP2340]

[dufeu]

[ct85711]

Well, going by what you have posted, I am assuming pyrogyro is the one that can ping the default gateway. The interesting part, I am specifically seeing, is that the 3 machines have different routing tables.

pyrogyro

[LIsLinuxIsSogood]

[eccerr0r]

Promiscuous mode is normal and required on soft bridges.

As for the question on hand I still don't get how the interfaces got configured the way they did - they look very weird. Perhaps best to rip up all the network configuration and start over - configuring all the same way, unless some details on how they were configured would be helpful.

BTW, yes, if you have your netmasks wrong, it's still a toss up whether your network...works. You'll get lucky if all of them just so happen to have IP addresses on the same subnet anyway despite having incorrect masks. It's just when getting out of the subnet when things start breaking, and yes, there is no reason to have incorrect netmasks... don't do it!
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[LIsLinuxIsSogood]

Reset the default gateway on each of them to the router...that could really help a lot. I noticed at least one of the three machines doesn't have one set. Then what it does is probably forward all those packets to some other PC, or randomly assigned or something and then the other PC has to reply back to the non-working one to say, "hey there I'm no gateway router" so that won't be good. If only there were some automatic way to configure network settings, oh wait NetworkManager.

[SP2340]

[quote="dufeu"]

[LIsLinuxIsSogood]

[LIsLinuxIsSogood]

Why does pyrogyro and blaze have two interfaces, are they separate network interface cards, so you are actually attempting to have how many total addresses be able to "ping" one another...3 or 5 addresses?

EDIT: I take that back, looks like 4 NIC's, since on pyrogyro the interfaces share same ethernet address

Also, I can't tell what it is about blaze that isn't working -- other than a stale route table, the ping to router is working. What else are you trying to do other than that with it?

Since you've gone ahead and changed the subnets already, for pyrogyro, which was already working before you started making changes, I just want to point out the differences between the previous (old) and current settings...

[LIsLinuxIsSogood]

[ct85711]

First off, generally you should not ever have to mess with the arp table, as it is automatically updated on it's own. The way IPv4 works is that the computer will send a arp request out (a broadcast) onto the network, asking who has that IP address. The computer has that IP address will respond back with it's mac address, which is added to the first computer's arp table. Then it sends the packet of data out. As the age grows (since the last time it communicated to that mac address), the system will first mark it as STALE, meaning it is old. After a while, the system will automatically clean out old entires from the arp table (it also waits for a minimum amount of entires too, to minimize wasting cpu time).

Beyond that, the arp table is generally volatile, meaning when you restart your computer it is wiped out and nothing is saved. Having a static arp table entry can cause you issues if/when the network changes (as the system won't know of the change nor try finding out if it has changed on those ip addresses). The routing table on the other hand is saved. In generally, you should have at least 3 entries in it. First you will have one for the loopback, 127.0.0.1, next you will have the general network one for the subnet that computer is on. Last is the default gateway, in that anything that isn't on your network (which is based off your subnet mask using a bit-wise multiplication of the subnet mask and the computer's ip address), is sent there.

In this case where your subnet masks were incorrectly setup, has caused your arp cache to be corrupted (as the computer thinks it knows how to get there, but has the wrong information. So as LIsLinux suggested and verify everything have the correct subnet mask first; then flush the arp tables (or restart the devices). Otherwise I highly recommend you DO NOT add any static arp entries yourself, let the system manage the arp tables on it's own.

[eccerr0r]

Bridges have been notoriously annoying to set up, do the other two machines work if the machine with the bridge completely disconnected from the network?

And agreed, I don't see a reason for needing static ARP in this network configuration, it should work just fine (I think my network is similar... I have a machine with a bridge for VMs and a whole mess of other machines connected to switches, hubs and another bridge in pfSense...)
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[LIsLinuxIsSogood]

[eccerr0r]

You know, it's kind of funny I asked about the exact software configuration of all the machines on the OP's network, it is quite important as it would explain why the routing tables look so different to each other. If they were all configured by hand, that would explain that.

I don't know how much bridging experience people have, but a badly configured bridge can be an explanation for ARPs getting eaten. As they are in promisc mode, they are processing every packet that comes by and if they happen to respond to a packet they aren't supposed to respond to, it will cause weird things to happen like what we see here. It's worth an experiment to remove the bridge temporarily and reintroduce it to the network once others are fixed. We know bridging SHOULD work, and my example proves that point, but there's too many variables here. Taking one out will simplify things and help with debug. I'm not sure what circumstance would generate state "INCOMPLETE" but "FAILED" means no responders to an ARP and is the usual response to a missing machine.

Are there any firewall/iptables rules that could have masked some of the ARP responses too?

Also another thing is if possible reboot the switch, the switch does save state though by now it should have cleared...
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[SP2340]

[LIsLinuxIsSogood]

SP2340, this is not a matter of disagreement, so let's just wait for the OP to return with the progress on all three machines.

[dufeu]

[dufeu]

[eccerr0r]

Yes, bridges are supposed to be in promiscuous mode.

So they can ping each other, just nothing outside of the switch?

Technically all are on the same network as you have everything on 192.168.1.0/24. Just that these are all on switch1.

Switch1 (brand? model?):
PC1: br0:192.168.1.200/24 - can ping 192.168.1.1
PC2: eth0:192.168.1.201/24 - problems with 192.168.1.1
PC3: enp9s0:192.168.1.202/24 - problems with 192.168.1.1
Uplink...to somewhere...

Remote receiver of uplink... :
192.168.1.1/?? - router

You have an extensive network here it seems? How many switches/bridges?

You definitely have a weird problem here. Did you update the kernel recently as well? I can't think of any user space programs minus perhaps iptables that could do something like this. Tried flushing all your iptables/... firewall configuration, if you have any, just to be sure, including router?

What if you change the ip address from 201 / 202 to something else, maybe 21 / 22 if you're not already using them? And unplugging the working pc with the bridge didn't do anything? swapping ports? physically connecting to elsewhere on the network?
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[Ralphred]

[LIsLinuxIsSogood]

[LIsLinuxIsSogood]

[dufeu]

[eccerr0r]

It does not make sense any userland programs, except those that influence the kernel like iptables, could cause such behavior.
Try