solved my case of wan disconnect red light (it was not the FW)

[ugandy]

I've been struggling with this for a couple months now. hope this info is useful to others.

with ac88, ac86 and ax88
with 384.14 .15 and .16

plus one visit from comcast to improve line signals on their cable coming into the house.

and the problem would keep coming back:

ping times would go from 10ms to 4000ms, then downstream died with total packet loss, and wan light went red.
this could last minutes or hours.only on some random days!

Finally, remembering a comment here, about isolating possible factors, i realized i never considered that a client could cause this behavior.

So i mac filtered all clients and added them one by one.

Eventually found the culprit in my daughter's laptop
She'd been uploading her work video editing jobs and saturating the 5Mbps upstream.

google/snbforum to the rescue and I learn about:

bufferbloat
saturated upstream kills downstream (downstream needs to send packets up to regulate stream too)
no budget for WAN overhead, etc

I enabled QoS bandwith limiting option only, selected budget for WAN overhead, and set her upload limit to 1Mbps.
problem solved.

Cause of the wan red light in my case:
QoS disabled (200Mbps up / 5Mbps down; we never struggled for down bandwith so i never thought about enabling qos)
and
saturated upstream killed downstreams.




Some questions i'm left with:

1) It appears that the qos bandwidth monitor page, (and i think network monitor meters too), didn;t catch client saturating the line
until qos bandwith limiter enabled:
prior to enabling QoS i was seeing upload speeds of 200KB,
after enabling QoS, the speed readings became correct (1Mbps during her uploads, the limit i set).
the fact that i couldn't see that she was uploading while everybody else stalled, prevented me from investigating her client sooner. or maybe i missed it.

2) qos bandwidth monitor does not include spdMerlin usage, this is likely expected since it is not bound by QoS? during spdMerlin speed tests i noticed that clients downloads almost pause.

3) why didn't this happen on my old days of torrent clients? i didn't limit my torrent client upload speed then. is it a newer firmware problem? more clients on the routers?


4) if i change the qos from limiting to adaptive, should I install FreshJR? and I assume that adaptive is better than traditional?


thanks
hugo

 

[Deleted member 22229]

Problem is many others have had this happen when there upstream was not even in use or the connection itself for that matter. So if it indeed fixed it for you thats good but i fear in time it might likely happen again. I only had this happen once and my connection was not even being used. Nothing would resolve it except factory reset. Strange indeed. Has not happened again so far.

 

[ugandy]

Problem is many others have had this happen when there upstream was not even in use or the connection itself for that matter. So if it indeed fixed it for you thats good but i fear in time it might likely happen again. I only had this happen once and my connection was not even being used. Nothing would resolve it except factory reset. Strange indeed. Has not happened again so far.

before i enabled QoS, the webgui was reporting my upstream as using only 200KB, when in fact my 5Mbps was saturated. in my case this obfuscated the problem.
once i disabled both wifi radios and connected directly to a ethernet port, the problem went away and allowed me to continue the investigation. eventually setting a limit for this one client. this one was my particular issue and i'm hoping the red light doesn't come back

 

[ugandy]

why didn't this happen on my old days of torrent clients? i didn't limit my torrent client upload speed then. is it a newer firmware/ faster hardware problem? faster speeds?

if i change the qos from limiting to adaptive, is FreshJR a must?


thanks
hugo

any pointers on these 2 questions?
thanks

 

[dave14305]

I’m making a guess, but before QoS you might have had full NAT acceleration enabled and therefore not all packets would be accurately measurable. Once QoS is enabled some or all NAT acceleration is disabled and the router has better visibility to all traffic. I could be wrong, but at least someone can come and refute my theory.

 

[L&LD]

I'm here to refute that @dave14305 could be wrong! Word!