Re: 47% packet loss

Cihan · ‎2020-05-23

I've RBK53 where RBSatellites are connected to RBRouter with ethernet backhaul. Wifi covers whole house & I can hit my bandwidth limit which is 250mbps everywhere but currently experiencing an issue while gaming; to much lag/ms/latency or video conferencing.

RBRouter is in router mode and connected to my ISP's modem which is in bridge mode. I'm running latest firmware V2.5.1.16 on all devices

I have ran pingplotter to see where is the loss and below are the results.

As seen on the pictures, I've ~47% package loss on 192.168.1.1.

Does that mean I've the loss in my local network? In router or satellites?

Thanks in advance.

CrimpOn · ‎2020-05-29

@Cihan wrote:
I've installed the replacement unit in AP mode as well and there was no package loss. Then I changed it to Router mode & bridged my modem today. Packet loss on 192.168.1.1 is back.

Is this normal, anyone else is experiencing the same thing?

Attached is a Ping Plot I just did to Google.com. Zero packet loss to any node that responds to ICMP (notice that node 2 does not).

The Orbi has been replaced, but the problem remains and shows up only in router mode. I wonder:

The Orbi router will still have an IP address, even when in Access Point mode.
Can you run Ping Plotter against that IP address by itself?
It appears to me that the only "constant" in the problem is the Mac that is running Ping Plotter.
Can you get access to some other computer? (invite someone over with a laptop - and a mask, of course)
Would be useful to see if other devices report the same results.

If packets are actually being "dropped", I would expect the reports for every node after the Orbi to also report packet loss. Perhaps my understanding of how "ping plotters" function is incorrect. PingPlotter provides a really slick description of the product here:

https://www.pingplotter.com/fix-your-network/getting-started/how-pingplotter-works.html

But, they never say how those intermediate "hops" are discovered and measured.

My understanding is that the program does a "Trace Route" to determine how many hops there are and the IP address of each hop.

https://support.microsoft.com/en-us/help/314868/how-to-use-tracert-to-troubleshoot-tcp-ip-problems-i...
It then looks up the DNS names of those IP addresses to create a more "user friendly" display.

Then, it sends a series of ICMP requests to every hop in the router. i.e. to hop 1, to hop 2, to hop 3, etc.and records how long it takes for each response to return.

Something is going wrong with ICMP to hop 1, but not to any other hop. If the Orbi is "dropping packets", then it should drop packets that are going everywhere, not just drop packets intended for the Orbi router.

Sorry to go on (and on). I have a feeling the problem may be with the measuring process, not with the network.

View solution in original post

CrimpOn · ‎2020-05-23

I believe "in line" images have to be approved by a moderator before we can see them.

Perhaps you could "attach" them instead?

Cihan · ‎2020-05-23

this is google.com

schumaku · ‎2020-05-23

Which topology is used on the Ethernet Backhaul as per the What is Ethernet backhaul and how do I set it up on my Orbi WiFi System? ?

If there is a switch in the set-up - i this a non-managed one, or are we facing some more decent one(s) having STP enabled by default?

Cihan · ‎2020-05-23

First port of the router is connected to the unmanaged switch
Second & Third ports are feeding sat1 & sat2
They all are connected with cat6 cables.

Cihan · ‎2020-05-23

also the used switch is this one;

NETGEAR GS308-300PES 8-port Gigabit Switch

Cihan · ‎2020-05-23

This is for blizzard servers.

FURRYe38 · ‎2020-05-23

What are the ping results with a wired PC connected directly to the ISP modem with out the RBR in between?

Be sure your using the ping program with a wired PC to accurately test with.

Cihan · ‎2020-05-23

It looks like when you put Orbi in router mode, it goes crazy.

After changing modem into router mode and putting Orbi into AP mode fixed all the problems. So that means I cannot use Orbi in Router mode unless it is fixed with a new firmware.

So what are the differences & pros/cons of running Orbi in AP vs Router mode?

CrimpOn · ‎2020-05-23

@Cihan wrote:
As seen on the pictures, I've ~47% package loss on 192.168.1.1.
Does that mean I've the loss in my local network? In router or satellites?

Thanks for providing the images (which I see have now "appeared" in the original post). As an experiment, I downloaded PingPlotter for Windows on my PC which is connected directly to the Orbi RBR50. My results with Google.com are attached. I have zero losses to anywhere, and the latency to the Orbi averages 0.8ms. My Orbi is in router mode connected to a Spectrum cable modem (which is "only a modem").

The only obvious differences are (1) I am currently on firmware 2.5.1.8, and (2) my PC is connected directly to the router and there is no switch in between. The difference in firmware release might be responsible (?), which dampens my enthusiasm for updating firmware. I do not have a Netgear unmanaged switch (I have a couple of TP-Links, and will try connecting through a switch to see if that changes anything.)

Since your "problem is solved", there is little incentive to change the modem and router and run Ping Plotter when directly attached to the Orbi. I, however, am not confident that the Orbi being in router mode is solely responsible for the issue. There might be configuration differences between our Orbi's. It might be load related (I have only about 30 devices and most of them are "nearly quiet" most of the time.

Cihan · ‎2020-05-23

I get 0 ms on my ethernet connected pc as well so i guess my 3.2 ms average is caused by wifi.

Other than that my Orbi satellites are not behind the unmanaged switch, they are directly connected to RBR50. One of the ports of RBR feeds the switch which feeds solar panel unit, alarm system (no wifi devices or ethernet backup).

I have 22 clients up all the time and could be more when some stuff turned on. Doorbell, Arlo base unit, lots of active IOT devices running at the same time. I don't know if they are making this conflicts.

My ISP had suggested to restart the modem and it have happened when they have changed it from bridge mode to router mode remotely, i don't know may be it was related to modem?

Amazon has sent out a replacement and I'll be checking it out on wednesday if it behaves same. I'll do my tests without upgrading the firmware so it might give us a clue. If everything would be normal with the oob firmware, then I'm gonna try it with latest firmware to see if it is happen or not which will answer also if it is a hardware issue or not.

Btw I'll do my tests after removing switch from the network, to eliminate this parameter. If everything goes ok, I'll add it and test again to see difference.

Once it is delivered & tested, I'll let you know the results so it will give you a clue if it is a bad firmware or not.

Thanks for your interest.

CrimpOn · ‎2020-05-23

Thanks for the additional information. So, the "test path" is computer ->WiFi->Satellite-> ethernet backhaul link->router->modem, correct?

I will reload PingPlotter on a laptop and connect to my satellite over WiFi.

CrimpOn · ‎2020-05-23

I am getting no closer to a solution. PingPlotter on a Lenovo laptop -> WiFi -> satellite -> WiFi backhaul-> Orbi router. Results for Google.com are attached. Average round trip is under 6 ms compared to 0.8 when wired directly to router. Zero packets lost.

I struggle to explain dropping 47% of packets. I would love to see a few pictures of the CPU load on the satellite and router during the ping. http://<ip of sat or router>/debug.htm

Cihan · ‎2020-05-24

I'll rollback my change (de-brigding modem) and put Orbi to router mode to see if i can reproduce the case, otherwise I'll be waiting for the new unit to test.

Cihan · ‎2020-05-29

I've installed the replacement unit in AP mode as well and there was no package loss. Then I changed it to Router mode & bridged my modem today. Packet loss on 192.168.1.1 is back.

Is this normal, anyone else is experiencing the same thing?

PS: The new router has come with 2.3.30 or something firmware and my devices keep hopping and was not stable at all. I was frusturated and updated to the latest firmware. BTW unlike my first unit, it was not detecting the new firmware and I had to update it manually.

Since i would like to use router features (armor, disable internet via app etc..), AP mode is not a choice for me.

Cihan · ‎2020-05-29

btw I've discarded switch and only 2 satellites & security system were connected with wired ethernet. So unmanaged switch is out of the scope now.

CrimpOn · ‎2020-05-29

@Cihan wrote:
I've installed the replacement unit in AP mode as well and there was no package loss. Then I changed it to Router mode & bridged my modem today. Packet loss on 192.168.1.1 is back.

Is this normal, anyone else is experiencing the same thing?

Attached is a Ping Plot I just did to Google.com. Zero packet loss to any node that responds to ICMP (notice that node 2 does not).

The Orbi has been replaced, but the problem remains and shows up only in router mode. I wonder:

The Orbi router will still have an IP address, even when in Access Point mode.
Can you run Ping Plotter against that IP address by itself?
It appears to me that the only "constant" in the problem is the Mac that is running Ping Plotter.
Can you get access to some other computer? (invite someone over with a laptop - and a mask, of course)
Would be useful to see if other devices report the same results.

If packets are actually being "dropped", I would expect the reports for every node after the Orbi to also report packet loss. Perhaps my understanding of how "ping plotters" function is incorrect. PingPlotter provides a really slick description of the product here:

https://www.pingplotter.com/fix-your-network/getting-started/how-pingplotter-works.html

But, they never say how those intermediate "hops" are discovered and measured.

My understanding is that the program does a "Trace Route" to determine how many hops there are and the IP address of each hop.

https://support.microsoft.com/en-us/help/314868/how-to-use-tracert-to-troubleshoot-tcp-ip-problems-i...
It then looks up the DNS names of those IP addresses to create a more "user friendly" display.

Then, it sends a series of ICMP requests to every hop in the router. i.e. to hop 1, to hop 2, to hop 3, etc.and records how long it takes for each response to return.

Something is going wrong with ICMP to hop 1, but not to any other hop. If the Orbi is "dropping packets", then it should drop packets that are going everywhere, not just drop packets intended for the Orbi router.

Sorry to go on (and on). I have a feeling the problem may be with the measuring process, not with the network.

Cihan · ‎2020-05-29

Thanks for your reply @CrimpOn .

Pingman says The only hop that matters is the final destination. If you're happy with the latency and packet loss being seen at the final destination, then none of the other hops matter.

So as you said it is a matter about the application for now.

But I was already experiencing a worst loss on my gaming pc ~97 last week. It is a windows pc and i was doing tracing via winMTR, also was getting ~10-20 loss on my developer machine via cable connected. But today I'm checking them side by side and they all look good.

The only difference is,

Circle is not installed on this router
Daisy chain disabled
2.4 & 5 ghz channels are set differently.

Now I don't experience any issue at all.

Thanks for your help @CrimpOn !