Traffic spike fix from 384.14_2 was reverted in 386.1

[RMerlin]

Did they ask you to tell us to email them or did you just suggest that? If it was the former, you’d think they would expect emails. If it’s the later, I’m not sure they’ll ever respond.

They asked me to have people with that issue contact them through that email address, and also specify in that email that I referred them to that email address. That's all I can tell.

 

[Tech9]

Hmm... I've never seen this bug on AC86U, but I have now one AC66U B1 and looks like I have accidentally downloaded the entire Wikipedia in few seconds time. Impressive. This router is running for 2 days only and that happened really quick.

 

[Morac]

Yep that happens on my AC88U for a few days after it restarts, then the spikes go away. That said I've found that the accuracy even without the spikes has gone down. With the 384 firmware, the monthly traffic was usually around 5% higher than what Comcast measured. With the 386 branch, it's 10 to 15% lower, even for months without the spike. If there's a spike all bets are off.

For example in March Comcast showed I used 994 GB of data, while the router shows 868 GB, a difference of 12.6%. For February Comcast shows 429 GB and the router shows 280 GB, a 34.7% undercount.

 

[Jeffrey Young]

Yep that happens on my AC88U for a few days after it restarts, then the spikes go away. That said I've found that the accuracy even without the spikes has gone down. With the 384 firmware, the monthly traffic was usually around 5% higher than what Comcast measured. With the 386 branch, it's 10 to 15% lower, even for months without the spike. If there's a spike all bets are off.

For example in March Comcast showed I used 994 GB of data, while the router shows 868 GB, a difference of 12.6%. For February Comcast shows 429 GB and the router shows 280 GB, a 34.7% undercount.

To my understanding, any traffic that does not go through the kernel will not get counted (Flow Cache for instance). I think VNstat has the same issue.

 

[Morac]

To my understanding, any traffic that does not go through the kernel will not get counted (Flow Cache for instance). I think VNstat has the same issue.

I think all of it should. I never enabled the Trend Micro stuff on my router.

 

[anotherengineer]

I would assume Johns fork is unaffected by this?

 

[DeepAnger]

I would assume Johns fork is unaffected by this?

It is (unaffected) on my RT-AC68U.

 

[john9527]

I would assume Johns fork is unaffected by this?

It's literally been years, but I remember that my fork had the problem as well and spending a lot of time trying to track it down. I could never find the root cause and it was always just a single, sporadic data point, so finally just wrote a fix to filter out the bad data point by duplicating the previous data point.

 

[JackJack]

Using 386.2_6 on a RT-AC86U
Found how to recreate spike issue.

Since a few releases, those spikes are few and probably I've found a scenario where there's a high probability to recreate a spike.
Early 386.x firmiwares, all you had to do for those spikes was just waiting (usually few minutes).

Now, after some checks, I've found that this spikes still happen when there's a WAN down/up status.
To recreate, just click on "Internet connection" switch inside "internet status" to disable LAN. Then after some seconds switch it on again.

That is about the same as having internet connection going down for any other reason.

Half the time you'll get a traffic spike on "traffic analyzer" that can be seen in "last 24h". Also "daily" section will report many gigabytes of traffics. Just as an example, in one day I got reported almost 1Terabyte of data exchanged. That is impossible by some order of magnitude.

 

[Tech9]

Using 386.2_6 on a RT-AC86U
Found how to recreate spike issue.

Not the same spike issue this thread is about.

 

[DeepAnger]

I would assume Johns fork is unaffected by this?

I've briefly tried 50E8 and did not see them (the spikes). But 386.3 still has them. That is what appears in less than a few minutes.

My main router (RT-AC68U) now runs the latest FreshTomato.
I've bought two routers since last year and these are not from ASUS. I think it's called vote with your wallet.

 

[RMerlin]

I've bought two routers since last year and these are not from ASUS. I think it's called vote with your wallet.

The issue was fixed months ago by Asus in all of the newer models. Only the older models still have that issue. When neither myself or them were able to reproduce the issue on these older models a public call was made for people experiencing the issue to contact them. It didn't happen. So there we are now.

 

[Morac]

The issue was fixed months ago by Asus in all of the newer models. Only the older models still have that issue. When neither myself or them were able to reproduce the issue on these older models a public call was made for people experiencing the issue to contact them. It didn't happen. So there we are now.

i contacted them, they just never responded. Fortunately on older models, the problem appears to go away after a few days of uptime.

 

[DeepAnger]

The issue was fixed months ago by Asus in all of the newer models. Only the older models still have that issue. When neither myself or them were able to reproduce the issue on these older models a public call was made for people experiencing the issue to contact them. It didn't happen. So there we are now.

I did not want to appear bitter here. What did not happen was ASUS fixing the problem or at least replying to my e-mail.
The 386.3 was a clean install. I have VDSL 23 Mbps / 2.5 Mbps and I often download from Usenet. The RT-AC68U wan port is connect to the "livebox" and the first LAN port is connected to a CISCO switch. As soon as I start a download these spikes appear. It's hard to believe that ASUS can not reproduce this.

 

[JackJack]

I've started to use vnStat (entware) and this software can calculate traffic as expected and isn't influenced by those spikes.
So it's not a kernel issue.

On my setup those spikes can be easily reproduced by turning off WAN and then back online. Doing so there's high probability (more than 50%) to get a spike withing a minute. (from GUI, network map, internet status, internet connection off/on)

I had sent a mail to asus as well.

 

[Kapricornus]

Also reported to Asus. I still have the issue with Upload Stats on AC-66U B1 and 386.3_2 firmware.
Usually it happens after I upload 2-4 Gb worth of data