mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Linux (https://www.mersenneforum.org/forumdisplay.php?f=39)
-   -   NIC gone nuts (https://www.mersenneforum.org/showthread.php?t=25546)

Prime95 2020-05-15 03:04

NIC gone nuts
 
Every now and then one of my Linux boxes goes nuts. It must be flooding the network because it takes down the whole local network until the ethernet cable is unplugged.

Could this be a Linux issue? If not, are there any ideas for a software fix to workaround the issue?

paulunderwood 2020-05-15 03:40

[QUOTE=Prime95;545414]Every now and then one of my Linux boxes goes nuts. It must be flooding the network because it takes down the whole local network until the ethernet cable is unplugged.

Could this be a Linux issue? If not, are there any ideas for a software fix to workaround the issue?[/QUOTE]

I had a "crazy" network recently. It turned out to be a half-in LAN cable on one of the machines on my network.

chalsall 2020-05-15 18:12

[QUOTE=Prime95;545414]Could this be a Linux issue? If not, are there any ideas for a software fix to workaround the issue?[/QUOTE]

It could be any one of many things.

Suggestion: Capture network traffic from at least one Linux boxen in addition to the suspect one. Use [URL="https://www.wireshark.org/docs/wsug_html_chunked/AppToolstcpdump.html"]tcpdump[/URL] to write the network traffic to a log file.

Then when the issue manifests, examine the logs for the period in question. This will tell you what type of packets are being broadcast, and potentially by whom.

Also, as Paul said, it could be a hardware issue -- something as simple as a bad cable can cause problems. Ditto switch loops in more advanced topographies.

kriesel 2020-05-23 02:13

Multiple links between switches must be defined as trunk lines or as primary and failover, or STP in place, to avoid traffic storms. Bridging/switching loops [URL]https://www.techrepublic.com/article/bridging-loops-and-spanning-tree-troubleshooting-in-a-cisco-switched-environment/[/URL]
Broadcast storm [URL]https://www.techopedia.com/definition/6270/broadcast-storm[/URL]

NIC hardware failure [URL]https://unix.stackexchange.com/questions/83273/how-to-diagnose-faulty-onboard-network-adapter[/URL]
I've seen it where a failing NIC brought a LAN to its knees when that system was undergoing backup or other traffic.

Network software misconfiguration on a node. I knew someone who (mistakenly) configured routing on a 1-interface slow VMS VAX on a Friday. It made enough of a mess she got two departments' LANs severed from the campus backbone for the weekend by Network Operations.

Software running can create the appearance of a hardware issue. A coworker opened a malicious attachment and then left his office for a meeting in another zip code. The attachment launched his computer as a spam server to the world, saturating the department WAN connection to the point of unusability, until his system was identified as the source and disconnected from the LAN.

Violating network layout rules (such as in 10Mhz networks, too many hubs chained; a media converter counts as half a hub) can cause trouble such as fragmented, runt packets and error counts.
Optical or electrical attenuation in the links; stuff gets jostled or damaged, and ages.

A managed switch with port monitoring & mirroring capability can be useful to have around either already installed or to insert when trouble is occurring. The port with an incipient NIC hardware failure connected will over time show very different statistics than others. Volume, errors, fragmented packets, etc. These can be very economical (<$50 used)
[URL]https://www.ebay.com/itm/Dell-F491K-PowerConnect-2824-24x10-100-1000-2-SFP-Web-Managed-Switch/113956592097[/URL]
[URL]https://downloads.dell.com/manuals/all-products/esuprt_ser_stor_net/esuprt_networking/esuprt_net_fxd_prt_swtchs/powerconnect-2824_user%27s%20guide_en-us.pdf[/URL]
(Not a model recommendation, just the first I found. Note the cable fault test on this one. I used to buy 3com new. Managed switches were equipped with IP#'s just outside the address space of the department LAN, and management workstations equipped with broader netmasks to talk to the switches too.)

A laptop with a good network capture and analysis program installed can be helpful.
[URL]https://techtalk.gfi.com/the-top-20-free-network-monitoring-and-analysis-tools-for-sys-admins/[/URL]

A possibility for mapping and monitoring your network: [URL]https://mikrotik.com/thedude[/URL] or similar.

A PCI or PCIe NIC might be useful if an on-motherboard NIC is going bad. Disable bad one and install new NIC.

Prime95 2020-05-23 04:45

[QUOTE=kriesel;546263]A PCI or PCIe NIC might be useful if an on-motherboard NIC is going bad. Disable bad one and install new NIC.[/QUOTE]

I have 2 mobos doing this. My first plan of attack is to try configuring the Asrock h110mitx/ac wireless card to provide internet access instead of the NIC. Fingers crossed.

Prime95 2020-05-24 04:41

I finally had time to look at this. You think setting up a wireless card should be easy? Think again. After wasting a few hours, I've given up. I'll probably end up retiring the mobo instead.

henryzz 2020-05-26 12:37

[QUOTE=Prime95;546331]I finally had time to look at this. You think setting up a wireless card should be easy? Think again. After wasting a few hours, I've given up. I'll probably end up retiring the mobo instead.[/QUOTE]

I have used USB wireless sticks with no issues. I think they have been completely plug and play on windows and linux.

storm5510 2020-05-26 13:24

[QUOTE=Prime95;546331]...You think setting up a wireless card should be easy? Think again. After wasting a few hours, I've given up... [/QUOTE]

I have an extremely small one, like what a wireless mouse uses, which plugs into a USB port. It is [U]not[/U] plug-and-play. It came with a mini-cd which has drivers and software. I tried plugging it in both a Windows 7 and Windows 10 system, without the drivers. Neither indicated anything new. I would much rather, and did, run 50 feet of Ethernet cable as opposed to using it again. Sometimes it worked, and other times, it did not. It simply is not worth the effort.

henryzz 2020-05-31 17:57

[QUOTE=storm5510;546491]I have an extremely small one, like what a wireless mouse uses, which plugs into a USB port. It is [U]not[/U] plug-and-play. It came with a mini-cd which has drivers and software. I tried plugging it in both a Windows 7 and Windows 10 system, without the drivers. Neither indicated anything new. I would much rather, and did, run 50 feet of Ethernet cable as opposed to using it again. Sometimes it worked, and other times, it did not. It simply is not worth the effort.[/QUOTE]

Was the PC connected to the internet at the time?

Prime95 2020-05-31 19:01

For those still interested -- things have gone really, really weird....

The system in question is a bare bones mini-itx board (from the dream machine build) with a Radeon VII and minimal Ubuntu Server installed. Sitting right next to it is the exact same setup, sharing a power supply. This was necessitated when a Haswell mobo running 2 GPUs died.

Most of the time this setup works great. One system's NIC goes nuts every few days bringing the entire local network to its knees. The other system has never given a problem. Simple diagnosis -- bad NIC.

So I swapped the mobo with another dream machine mobo. The new mobo has the same problem. The old mobo back in the dream machine? Nary a problem.

So I change the ethernet cable and the switch it is plugged into. No difference.

My latest desperate attempt is a 50 ft' ethernet cable connected to the router connected to the cable modem -- bypassing as much local networking gear as possible. I don't have much hope this will make a difference.

If this fails, the only hardware that has not been replaced is the GPU and power supply. I'll swap the GPU with the identical system sharing the same power supply. How a GPU blip would cause a NIC problem that takes down the entire local network is a mystery to me.

EdH 2020-05-31 19:28

Currently, I have a headless RPi B nearly taking down an entire switch when I connect it. At first, I thought the switch went bad, but a replacement acted the same. Then I traced it to the RPi. I've not troubleshot it yet to see if I can tell why, but when I plug in the net cable, the rest of the machines on that switch lose any timely access within the LAN. The rest of the LAN continues as normal.


All times are UTC. The time now is 05:02.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.