DEC3920 Quick Review

[OPNenthu]

Yes, you're right.  They disabled it on the VP2440's I226-V interfaces, but none of the V-series ones were affected.

[dirtyfreebooter]

ugh, the drop out happened 2x times tonight. this is turning into a disaster. i've emailed decsio for support. hopefully they have some answers.

[Seimus]

Hm,

Sorry to hear you see such problems. If there is no logs present I would suspect its in firmware related problem.

Hopefully Deciso can provide some solid feedback.

But what you describe sounds like the Issue ppl had on i226 around firmware 2.17 which was related to ASPM & Power savings.
https://forum.opnsense.org/index.php?msg=247197

Regards,
S.

[dirtyfreebooter]

response from deciso

There's a lot of misinformation about the Intel nics unfortunately, on our end these have proven to be highly stable for years, currently we use the Intel 226, but the 225 versions we had before have been solid as well (using the right firmware and powered properly, poor hardware design choices in our experience cause quite some issues as well).
 
When the machine reports "host down" the first question you need to ask yourself is if the port is properly linked, our documentation contains some pointers about where to look in these cases, you can find it here https://docs.opnsense.org/hardware/support.html#network-connectivity-issues

Please make sure to test using default settings, no specific tunables whatsoever to prevent these causing unforeseen issues. You can always share the output of the commands in our documentation for us to have a small look, if you need an engineer to assist, that's also possible, but comes at additional cost.

i know they can't test everything, but i have used this config in so many varieties of OPNsense and it worked fine. Supermicro X11SCL-iF, Protectli VP2440, Odroid H4 Ultra, Lenovo P3 tiny, Aliexpress N200 topton and no issues.

its not a complex config and i didn't use any tuneables except from the shipping config and the 2 zenarmor adds. since friday i have had now 4 cases of networking goes down on i226, no logs, no output from kernel. NOTHING. unplug/plug the WAN cable and instant connection again.

i plugged my VP2440 back in last night and its now its been 12+ hours of working fine.

prior, my OPNsense uptimes are basically the OPNsense business release dates, and my internet, fiber, has also only went out 1 time is over 5 years and that was due to construction company cutting lines in the neighborhood.

my take at this point is that, Decsio $3000 firewall doesn't work and i have to pay to maybe figure out the issue with their firmware is sus. at this point, i guess my only other option is to send it back and pay for return shipping.

frankly what seems amazing, turned quickly in a horrible experience

[patient0]

You started the thread on Easter Friday (bank holiday) and it's now Easter Monday (bank holiday), give Deciso some time to react.

[dirtyfreebooter]

You started the thread on Easter Friday (bank holiday) and it's now Easter Monday (bank holiday), give Deciso some time to react.

i am aware of the holiday and i asked them to not reply during their break. i also have 30 days to return and US customs held the device for more than a week. i think i am entitled to some support of their $3000 firewall and basic DHCP WAN dropping connection silently. also to let others know my experience, since there is very little out there on what its like deal with an issue with official opnsense hardware.

--

i am more concerned with the initial response itself, basically telling me i226 is perfectly stable and should check my network cables (switch i already verified multiple machines (vp2440, supermicro x11scl-if) and multiple cables before i even bothered contacting them). i am not asking for help setting up my network, behavior points to some sort of hardware/firmware subtle bug is happening.

a better response, if any during holiday, was "its holiday in EU, we will be back on tuesday" vs gaslighting me on i226 issues and my network cable.

also, based on the response, quoted above, it seems unclear they really are going to even follow up. they basically told me i can pay for support or return it, paying for shipping back to netherlands.

[patient0]

What device is WAN connected to, an ONT? For testing purposes, can you add a small switch between the OPNsense WAN and whatever your ISP provided?

[dirtyfreebooter]

What device is WAN connected to, an ONT? For testing purposes, can you add a small switch between the OPNsense WAN and whatever your ISP provided?

ONT - Calix 711GX (outdoor, not a ONT/Router combo)

The OPNsense systems i've used over the past 5 years with the same ISP / same ONT:

• Supermicro X11SCL-iF with igb WAN and ixl LAN (X710-DA2 nic)
• Odroid H4 Ulta with igc WAN and LAN (i226v)
• Protectli VP2440 with igc WAN and ixl LAN (X710-BM2)
• Lenovo P3 tiny igb WAN and igb LAN (i350-t4)

none of these system exhibit the behavior of the DEC3920 with the WAN going out with nothing in dmesg or console. my ISP has only ever gone out when a construction company cut wires while putting in a new bridge in the neighborhood. i consider the ISP/ONT rock solid by itself, though understand it could be some issue with combination, specific to the DEC3920

i have some unmanaged 5-port switches. Netgear 1Gbps model and Netgear 2.5Gbps model. so i can do this test. What specifically am i looking to test by adding in the switch?

[patient0]

i have some unmanaged 5-port switches. Netgear 1Gbps model and Netgear 2.5Gbps model. so i can do this test. What specifically am i looking to test by adding in the switch?

Quite a few issues that come up here or on the other sense forum are between the upstream device and the router WAN. E.g. the upstream device is picky or is not as flexible as switch ports or a port flaps. Adding a switch would show if it is a specific issue of the two devices or not. Generally the routes WAN (and LAN) ports don't like the connected device ports flapping at all.

I'm aware that you have/had a number of devices that run well with the same upstream device and as a result the 3900 should behave well. The test would only be to narrow it down. It's of course also sensible to wait for help from Deciso.

[nero355]

i consider the ISP/ONT rock solid by itself

Can you rule out Firmware Updates done by the ISP ?
Has there been any recent downtime due to Network Maintenance during the night ?

Though understand it could be some issue with combination, specific to the DEC3920

To be honest I think it's simply the Intel NIC having issues since the whole i22x range had some seriously weird bugs over the years...

[dirtyfreebooter]

I'm aware that you have/had a number of devices that run well with the same upstream device and as a result the 3900 should behave well. The test would only be to narrow it down. It's of course also sensible to wait for help from Deciso.

yea, i will certainly see what deciso says this week, then i have to make a decision before the 30 days is up, with 10 days in US customs. so i am willing to gather as much information beforehand as i can and hopefully deciso doesn't ignore the effort.

[dirtyfreebooter]

Can you rule out Firmware Updates done by the ISP ?
Has there been any recent downtime due to Network Maintenance during the night ?

yea, this ONT is ancient 1 Gbps, i doubt it gets many updates. its been flawless for 5 years, literally, not joking or exaggerating. its impressive how stable this ISP is compared to other friends in town how only have Comcast cable and how much their connection goes down.

another note too, when the connection stops working, nothing in the logs / dmesg, but when i pull the cable and put it back in, connection comes back, but i also see the up/down event in the logs/dmesg, so nots 100% proof the ONT connection is stable, but some evidence it is, since i see the up/down events if i manually pull the cable.

i agree, i226, especially on FreeBSD with ASPM/power stuff seems sus. all the driver mailing lists posts i see seem to imply that that linux driver has specific code to prevent this TX hang / stall. any linux server, proxmox, truenas, i've run with i226 has never seen these types of network issues seen in FreeBSD/OPNsense.