Hello,

 

I'll preface this with last weekend I replaced some disks and had a couple of unclean shutdowns trying to figure some stuff out.  Currently the array is up and parity is valid again.

 

Now that parity is valid I wanted to dig deeper into my network issues and found that eth1 no longer exists on the network settings page, nor does bond0.  If I go to system devices both NICs (onboard, supermicro motherboard) are there.

 

I also cannot route outside of my LAN any longer from within unRaid.  nslookups work, but pings to domains or IPs do not.  I was bonding via LACP 802.3ad to a Cisco SG300 switch.  I changed bonding in unRAID to be active-backup instead and also switched ports on the switch.  All to no avail.  Running 6.8.1, downgraded to 6.8.0 thinking maybe something with the update did it and I didn't notice it but that didn't help either.

 

I also completely removed the network.cfg file and reset things up and still nothing - eth1 is just gone.

 

Thank You,

Josh

Edited January 23, 20206 yr by shwa87

Attached are diagnostics if it should help with this issue.

catan-diagnostics-20200123-1553.zip

@bonienl you've helped me out with some odd networking stuff. Any ideas on this one?

thanks @smdion.  I should also mention that the non working NIC lights do flash, and the port is up at gig/full on my switch.

First NIC isn't initializing correctly, second one is:

Jan 23 14:59:37 catan kernel: e1000e: Intel(R) PRO/1000 Network Driver - 3.2.6-k
Jan 23 14:59:37 catan kernel: e1000e: Copyright(c) 1999 - 2015 Intel Corporation.
Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.0: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
Jan 23 14:59:37 catan kernel: e1000e: probe of 0000:04:00.0 failed with error -2


Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.1: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.1 eth0: (PCI Express:2.5GT/s:Width x4) 00:30:48:62:4e:f1
Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.1 eth0: Intel(R) PRO/1000 Network Connection
Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.1 eth0: MAC: 5, PHY: 5, PBA No: 2050FF-0FF

No idea on the reason, try updating to v6.8.1, also look for a BIOS update if available.

3 minutes ago, johnnie.black said:

First NIC isn't initializing correctly, second one is:

Jan 23 14:59:37 catan kernel: e1000e: Intel(R) PRO/1000 Network Driver - 3.2.6-k
Jan 23 14:59:37 catan kernel: e1000e: Copyright(c) 1999 - 2015 Intel Corporation.
Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.0: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
Jan 23 14:59:37 catan kernel: e1000e: probe of 0000:04:00.0 failed with error -2


Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.1: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.1 eth0: (PCI Express:2.5GT/s:Width x4) 00:30:48:62:4e:f1
Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.1 eth0: Intel(R) PRO/1000 Network Connection
Jan 23 14:59:37 catan kernel: e1000e 0000:04:00.1 eth0: MAC: 5, PHY: 5, PBA No: 2050FF-0FF

No idea on the reason, try updating to v6.8.1, also look for a BIOS update if available.

I did notice that, too - I'm not sure why either - definitely not a bios update available, but maybe a re-flash?  I was on 6.8.1 when the issue started.

You can also try downgrading Unraid, different kernel, older or newer, might make a difference, if still the same with a release that was working previously there could be something wrong with the NIC.

Okay, so I figured out my issue of being unable to route out my LAN - my firewall had a static DHCP mapping to the mac address of the NIC that isn't working.  So now to try and figure this out.  I remade my USB drive and it still is behaving the same.  I can settle on only having 1 NIC, but this is stupid.

Well I tried a BIOS reflash and that didn't help.  It's baffling to me why the kernel driver loads for one but not the other:

 

root@catan:~# lspci -nnk | grep -iEA3 "(network|ethernet)"
04:00.0 Ethernet controller [0200]: Intel Corporation 80003ES2LAN Gigabit Ethernet Controller (Copper) [8086:1096] (rev 01)
        Subsystem: Super Micro Computer Inc Motherboard [15d9:1096]
        Kernel modules: e1000e
04:00.1 Ethernet controller [0200]: Intel Corporation 80003ES2LAN Gigabit Ethernet Controller (Copper) [8086:1096] (rev 01)
        Subsystem: Super Micro Computer Inc Motherboard [15d9:1096]
        Kernel driver in use: e1000e
        Kernel modules: e1000e

 

Are there any other diagnostic or troubleshooting steps I can try?  Or can I try and force the driver to load again?  To me it is also weird that the port links up, and the switch shows it up at gig/full.  But maybe that is normal behavior?

1 minute ago, shwa87 said:

Are there any other diagnostic or troubleshooting steps I can try?

Do you know of an Unraid version where it was working? If yes try that, if still the same NIC is likely a goner.

 

12 minutes ago, johnnie.black said:

Do you know of an Unraid version where it was working? If yes try that, if still the same NIC is likely a goner.

I know for certain it worked under 6.8.0, and am fairly certain it worked under 6.8.1 as well.  Have tried both, although I did not do 6.8.0 as a fresh USB build. In your opinion might that make a difference?

1 minute ago, shwa87 said:

In your opinion might that make a difference?

Very unlikely.

This is now resolved.

 

So I found this post regarding my NIC and I also realized it was also about my exact system board.  So I did what worked for them - powered down and unplugged.  Powered back up and BOTH NICs were dead with -2 error code.  I looked at the back of the machine and one of the ports lights were solid.  So I unplugged from power again, and pulled the cables out of both, powered back on. And BOOM - both eth0 and eth1 are present.  The (apparent) issue: BOTH of my patch cables.  Neither one would allow either port to link up after both eth0 and eth1 were back.  Replaced both, and I'm back in business on BOTH.  I am so happy.

 

Lesson: It's (almost) always a simple and (sometimes) stupid solution.