Traffic spike fix from 384.14_2 was reverted in 386.1

[palirobe]

As I said earlier. I cannot explain why changing the priority of ONE device worked for me, but it has. I've even rebooted my router as Morac suggested & I'm not getting any traffic spikes after reboot. This solution works for me until Asus can get a patch put in to fix the problem. It's highly unlikely it will work for everyone, though it may be worth a try.

signing off as palirobe
Senior Analyst
Computer Software Engineering Technology

 

[Crazy]

SOLUTION that worked for me: I'm on RT-AC88U. Traffic Analyzer on version 384.19 worked well. After upgrading to 386 & enabling QOS I saw enormous traffic spikes (about 17 billion gig every 3 hours). After blocking internet to my devices (one at a time), I discovered the offending device was my AirTV. AirTV is a device that broadcasts OTA antenna channels via the Sling App. The AirTV device errored/spiked whether it was a WIFI connection or a wired connection. My work-around was simple. I changed the AirTV to have the highest priority under QOS & the traffic spikes stopped. I'm not sure why they stopped, but this change worked in adaptive, traditional & bandwidth limiter options. Traffic Analyzer is working normally for me now. If others on this forum can find the offending device(s) on their routers, I'm hoping that changing the priority on the offending device will fix the traffic spikes for them as well.

I don’t have QOS on but mine have spike. I’ve also noticed on daily report sometimes it missing a few days.

 

[CaptnDanLKW]

My real-time monitoring scale is borked. When the scale is in KB, After navigating to the Traffic Analyzer screen the scale always shoots up (within 5 seconds) to a massive range and therefore the traffic graph is always flat-lined.

Settings the default scale to be in MB (and navigating away and back) seems to fix, which is odd.

Nevermind. The issue came back in less than a minute when the scale was set to MB

So In my case, realtime is busted too.

No QOS, No diversion, FlexQOS, Skynet.

AI-Mesh - 2x RT-AC86Us
HW Accel - Enabled


I tend to have the massive spike issue in the 'Last 24 hour' screen, but this time it's accurate.

 

[Morac]

I don’t think that’s the same issue. “Massive” spikes will show up as impossibly high amounts of traffic. I‘m talking billions of TB per second.

 

[palirobe]

I'm still baffled why changing the priority of my AirTV caused the traffic spiking to stop. Traffic Analyzer has been stable for about 1.5 weeks. I rebooted on the 2/27 to see if it was stable after a reboot and it is, that's why my attached file shows only 3 days of traffic. Never seen a weird-o bug like this on a router before, but what do I know. I've only been a programmer for 20+ years. Bug seems to affect people differently. Good luck. The priority change was the last thing I tried & I finally got the analyzer to work.

 

[Morac]

I’m still seeing normal traffic except for the few days after a restart. I still have no idea why.

 

[dev_null]

I've posted a new version of Vnstat on Merlin here: https://www.snbforums.com/threads/b...-monitoring-with-full-install-and-menu.70727/ . It is a fully scripted install now.

I can confirm no spikes on the tested devices (AC66U_B1, N66U and AX86U), and might be a solution until the Asus issues are resolved and rolled into Merlin.

 

[RMerlin]

Any of you has had the chance to test 386.2 Beta 2 for a while now to see if the recent change I did to netdev_calc() helped? There was a potential variable overflow that could happen in particular to users with a VLAN. No idea if it was tied to this specific issue however, or if it was another issue that I just fixed.

 

[Wistuplu]

Dear RMerlin,

AC86U with 386.2 beta 2.
The issue was never visible for me with Cake ON. (Maybe related to "hardware acceleration" being OFF etc).

I just tested with Cake OFF and I got the spike within a few minutes at most. I had a spike for 2.4GHz wifi and also for 5.0GHz. They were not at the same instant. Yet the figure in the table for the "maximum" was very similar. I have nothing capable of those speeds. I'm not aware I would have a VLAN. I have a Netgear 8-port switch plugged into the router, and maybe one device that is 1 Gbps capable.

So, the maximum value displayed in the 2.4GHz table and in the 5.0 GHz table were:
2.147.341 KB/s
2.147.483 KB/s

I'm probably the last one to realize this but 2,147,483,648 is 2^31.

Best regards
Thank you for the software
W.

 

[Wistuplu]

FYI, RT-AC86U_9.0.0.4_386_42210-g5edd47f_cferom_ubi.w seemed to solve the issue with my AC86U. No ghost spikes after approx. 20 hours, and plausible totals.

If I understand correctly, Merlin 386.2 is merged with GPL 386_42095. So, maybe for a future cycle ;-)

PS: I could also reproduce the problem on Asus 9.0.0.4.386_41994

 

[RMerlin]

I just tested with Cake OFF and I got the spike within a few minutes at most.

Ok, thanks for trying. I must have fixed a different potential value overflow issue then.

 

[Morac]

I’m still on 386_1, but my AC88U has been up for 40 days and I haven’t had a spike since Feb 16th. I had it for 4 days in a row after rebooting, but nothing since.

 

[RMerlin]

I’m still on 386_1, but my AC88U has been up for 40 days and I haven’t had a spike since Feb 16th. I had it for 4 days in a row after rebooting, but nothing since.

It only happen under a very specific scenario. In all of these years, I have never experienced it myself.

 

[Morac]

It only happen under a very specific scenario. In all of these years, I have never experienced it myself.

I can reproduce it simply by restarting the router as I did that several times, but that’s the only thing that will do it. I rebooted my cable modem recently and that didn’t trigger any spikes.

 

[ech]

I've upgraded my router to an AC86U in the meantime (from an AC68U) and haven't seen the spikes in the 24h or daily roll-ups any more, but do see them in the real-time view, within 5-10s of looking, it will jump to 2147500KB/s. But since it's not affecting the overall data, I haven't put in the time to test a beta release w/ kids in the house who enjoy consistent internet access.

 

[dev_null]

To those who experience spikes: can you take a look at the system logs during those periods and confirm that NTP had sync'd appropriately (that is, compare the hours in the graph where the spike shows to the time and status of NTP sync shown in the log)?

I saw something weird with my test router when it rebooted after a short power outage but failed NTP sync: a series of hours with multi-GB use reported even though nothing was connected to it.

 

[RMerlin]

Asus sent me a patch that they recently applied to the kernel itself (rather than rstats), and which they believe might possibly be related. I haven't had time to study it, but I intend to apply it for the next release (be it beta 3 or final, I haven't decided yet).

 

[RMerlin]

Please test the builds in the Kernel Patch folder here: https://www.asuswrt-merlin.net/test-builds

 

[Jack Yaz]

Asus sent me a patch that they recently applied to the kernel itself (rather than rstats), and which they believe might possibly be related. I haven't had time to study it, but I intend to apply it for the next release (be it beta 3 or final, I haven't decided yet).

i wonder if that patch will have the happy side effect of preventing traffic spikes in vnstat etc. too!

 

[RMerlin]

i wonder if that patch will have the happy side effect of preventing traffic spikes in vnstat etc. too!

Could be, since the patch is at the kernel level. Feel free to give it a try, I built all HND models with the patch.