Weird Google problem

[Rik Stigter]

Hello!

My setup is as follows:

• RT-AX88, running Merlin 384.15

1. amtm
   1. Skynet v7.1.0
   2. Scribe v2.4.2
   3. Unbound Manager v2.07
   4. UiScribe v1.2.1

• • Swap 2.0GB
• AiMesh RT-AC86, running Merlin 384.15

These systems have gone through several configs (from 384.12, before that the AC86 was the 'main'-router) and have had the same problem since early December 2019. The problem is that after a few hours (usually 2-4) connections to anything to do with Google (including fonts), becomes very slow and then drops out altogether!
Everything else works, that is all other websites load without problems, although I have the impression that they are a bit slower than usual.

If, for testing purposes at such a time, I start a VPN (NordVPN, in my case) on a workstation/laptop everything works as is to be expected: fast. After a router-reboot: the same, but unfortunately only for the same 2-4 hours.
There is one other weird phenomenon and that is that I sometimes loose 5Ghz-connections, but they re-establish themselves if (big IF) I'm patient. I'm using both WiFi5 and WiFi6-adapters and all show this behaviour.

I have (diligently, if I may add), reset (M&M) both systems several times now, usually when upgrading the firmware. Before Unbound I used Diversion, but the problem persists no matter what.

My (I hope wrong ) ideas/conclusions so far are:

1. There's 'nothing' wrong with DNS, because all other websites load as to be expected
2. Memory is not an issue, for the same reason (it usually settles at about 80% max)
3. All other performance indicators look normal (no CPU-hogs)
4. There must be a Google-killer Gremlin somewhere (I'm not a great fan of Google myself, but I don't hate them or something. The thing is that my better half uses Google all the time and I don't want her upset!)

I am considering re-establishing the AC86 as my main-router to test whether it is the AX88 that is the culprit, but before I do that I thought I had better ask the bright spirits here if they have any better ideas than mine.
(Worse even: I'm considering flashing he AX88 with stock firmware. Huh.)

My ISP is showing an interest and they are sending me a new modem (well: router in bridge mode) which will arrive in a day or two. They even know about merlin-firmware (Hooray! Miracles do happen) and started looking at these pages when I called (even better). But they didn't come up with any better ideas than that, I'm afraid.

What is it that I'm overlooking?

 

[ColinTaylor]

1. There's 'nothing' wrong with DNS, because all other websites load as to be expected

I suspect there is something wrong with DNS. My guess would be that it's overly aggressive ad-blocking. You say you are no longer using Diversion, are you now ad-blocking in Unbound?

Regardless, of the four customisations you have made to the router I would start by disabling Unbound and seeing if that is the cause of the problem. If it isn't then try disabling Skynet. As this problem only takes a few hours to manifest it should be an easy test.

 

[Rik Stigter]

I suspect there is something wrong with DNS. My guess would be that it's overly aggressive ad-blocking. You say you are no longer using Diversion, are you now ad-blocking in Unbound?

Regardless, of the four customisations you have made to the router I would start by disabling Unbound and seeing if that is the cause of the problem. If it isn't then try disabling Skynet. As this problem only takes a few hours to manifest it should be an easy test.

Yes I am, but why am I the only one reporting this problem? My black- and whitelists are stock.
But will disable to test! Thank you!

 

[L&LD]

This does sound perplexing!

Curious why you don't use Diversion (and more importantly, let it install Entware for you)? Even if you disable Diversion after it is installed, I feel the system is much more stable for the many installs I've performed on various Asus supported routers vs. just installing Entware 'bare'.

Is the USB drive warm/hot? Putting it on a short extension cable may help.

Is the USB drive running in USB 3.0 mode? You may want to switch that to USB 2.0 mode, at least to test for that particular drive.

Have you tried something like this?

https://www.snbforums.com/threads/amtm-the-asuswrt-merlin-terminal-menu.42415/page-86#post-549970

Is it easy to test with just one/specific devices on the network until the issue re-appears?

Does this issue only affect wireless clients or wired too?

Kudos on the M&M installs, but I have to ask. No saved config files re-imported afterward? No changing of default settings, scripts or other customizations carried forward 'because it worked previously'?

I see that @ColinTaylor may be onto something!

 

[AndreiV]

I have an Ac3200 , latest Merlin version, Google Earth became totally unusable for several days , my son described it as "looking at a cabbage leaf through a teabag".

Changing from Quad9 DNS on the router helped slightly, using Nord cured the issue instantly.

I tried resetting Earth, uninstall/install etc. nothing helped . Then I tried 1.1.1.1 DNS and on all but one device things improved greatly but it is very slow.

Change router and the problem is gone.

 

[Rik Stigter]

This does sound perplexing!

Curious why you don't use Diversion (and more importantly, let it install Entware for you)? Even if you disable Diversion after it is installed, I feel the system is much more stable for the many installs I've performed on various Asus supported routers vs. just installing Entware 'bare'.

Is the USB drive warm/hot? Putting it on a short extension cable may help.

Is the USB drive running in USB 3.0 mode? You may want to switch that to USB 2.0 mode, at least to test for that particular drive.

Have you tried something like this?

https://www.snbforums.com/threads/amtm-the-asuswrt-merlin-terminal-menu.42415/page-86#post-549970

Is it easy to test with just one/specific devices on the network until the issue re-appears?

Does this issue only affect wireless clients or wired too?

Kudos on the M&M installs, but I have to ask. No saved config files re-imported afterward? No changing of default settings, scripts or other customizations carried forward 'because it worked previously'?

I see that @ColinTaylor may be onto something!

I did use Diversion and the problems were the same but I switched to the blocking system in UnBound (as is more or less advised once you have installed that). I did install entware through Diversion-install. The USB drive isn't particularly hot, but I will (later, one should test one thing at the time my old prof would say) put it in USB2-mode. As for the URL you mention, my reaction would have been the same. But hey: I've seen similar things happen and sometimes things just don't make sense (at least at first).
It is possible to bring the network back to one workstation only.
All clients are affected. I did not use saved settings at all, since you advise(d) against that!

Thanks!

 

[Rik Stigter]

I have an Ac3200 , latest Merlin version, Google Earth became totally unusable for several days , my son described it as "looking at a cabbage leaf through a teabag".

Changing from Quad9 DNS on the router helped slightly, using Nord cured the issue instantly.

I tried resetting Earth, uninstall/install etc. nothing helped . Then I tried 1.1.1.1 DNS and on all but one device things improved greatly but it is very slow.

Change router and the problem is gone.

I suppose you mean 'change DNS-provider and the problem is gone'?

But anyway: this problem has occurred when I was still using other DNS resolvers like one.one.one.one, Quad9 or Google (I measured them all to test whether it was DNS related and at the same time that I could'nt load websites from them I had Google DNS resolve timings of > 2200 ms. That bad! I even asked my ISP whether they had (again, since that happened in 2017) capacity problems in the direction of Google (Google has a very large datacentre only 14 miles from here). They say no. I believe them (this time).

 

[AndreiV]

No, changing DNS server helped slightly, physically changing to an old Billion router and the issue was gone.

 

[Rik Stigter]

No, changing DNS server helped slightly, physically changing to an old Billion router and the issue was gone.

Ah, OK. So you say it is hardware-related?

It's too early for me to conclude that, but as I've said above, I might exchange the routers if I don't get any further with reconfigurations.
Thanks!

 

[Rik Stigter]

I did use Diversion and the problems were the same but I switched to the blocking system in UnBound (as is more or less advised once you have installed that). I did install entware through Diversion-install. The USB drive isn't particularly hot, but I will (later, one should test one thing at the time my old prof would say) put it in USB2-mode. As for the URL you mention, my reaction would have been the same. But hey: I've seen similar things happen and sometimes things just don't make sense (at least at first).
It is possible to bring the network back to one workstation only.
All clients are affected. I did not use saved settings at all, since you advise(d) against that!

Thanks!

I have one addition to make and that is that my notebook now tells me I have a connection to my network but no internet. But I do have that! I wouldn't be able to reply otherwise. Very strange...

 

[L&LD]

@Rik Stigter what do you have as your WAN DNS settings?

 

[PDinDetroit]

I would reset (M&M) both systems and leave them with bare minimum configuration (no add-ons). Then test this out to see if the issue resurfaces in that 2-4 hour range. Then, install add-ons 1 at a time and see if the issue resurfaces in that 2-4 hour range each time.

I think that could help narrow it down.

Question: the AC86U, is it wired back-haul or WiFi to the AX88U?

 

[Rik Stigter]

@Rik Stigter what do you have as your WAN DNS settings?

LAN-setting: DNSFilter is enabled and set to 'Router'. Global DNS set to 1.1.1.1, 1.0.0.1 and 9.9.9.9
WAN-settings:
Connect to DNS Server automatically No
DNS Server1 1.1.1.1 (have used several others as well)
DNS Server2 1.0.0.1
Forward local domain queries to upstream DNS No
Enable DNS Rebind protection No
Enable DNSSEC support No (RS. since UnBound)
Prevent client auto DoH Auto
DNS Privacy Protocol none

 

[Rik Stigter]

I would reset (M&M) both systems and leave them with bare minimum configuration (no add-ons). Then test this out to see if the issue resurfaces in that 2-4 hour range. Then, install add-ons 1 at a time and see if the issue resurfaces in that 2-4 hour range each time.

I think that could help narrow it down.

Question: the AC86U, is it wired back-haul or WiFi to the AX88U?

I agree, it could. But I've been there already. That being said, at that time I wasn't being thorough enough (didn't wait ca. 12 hours (I'd say that is the practical 'best' minimum) before the next step) so any conclusion from those tests is wacky. So I will do that again if my current test (no adblocking in UnBound) doesn't give me the result I would like.

The AC86 is back-hauled over WiFi.

 

[L&LD]

That should work.

Try this:

• Set the WAN DNS to Auto. Make sure you don't reboot your router in the next few steps.
• When the GUI is live again, click on the Network tab on the left and then the Internet icon which will show you the WAN/Internet Status page on the right.
• Copy those two DNS settings and go back to the WAN page.
• Set the WAN DNS to Manual and put those ISP DNS settings in.
• Test, test, test, test. Yes, be patient.

See if this makes any difference.

 

[Rik Stigter]

That should work.

Try this:

• Set the WAN DNS to Auto. Make sure you don't reboot your router in the next few steps.
• When the GUI is live again, click on the Network tab on the left and then the Internet icon which will show you the WAN/Internet Status page on the right.
• Copy those two DNS settings and go back to the WAN page.
• Set the WAN DNS to Manual and put those ISP DNS settings in.
• Test, test, test, test. Yes, be patient.

See if this makes any difference.

Done!
But for now that's it, too many variables otherwise and that would go against my old prof (Johan van Benthem) who was (lots!) younger then than I am now, but that only means I'm getting old... He gave 'Symbolic logic' BTW and we had a lot of fun together, especially when it came to the management of our college.

 

[ColinTaylor]

I have one addition to make and that is that my notebook now tells me I have a connection to my network but no internet. But I do have that! I wouldn't be able to reply otherwise. Very strange...

You see that message when Windows a) can't connect to dns.msftncsi.com, or b) download a test file from www.msftncsi.com.

This is sometimes seen when people try to block Microsoft telemetry and include either of the above addresses.

 

[Rik Stigter]

First test-results:

1. After > 12 hours the connection to Google seems to be stable when I disable Ad-Blocking in UnBound
2. Yesterday evening (I'm on UTC + 1 so about 8 hours ago) my tablet wasn't making a connection to Google. That may have something to do with another problem, so I'm going to store that info for a while (I have seen some strange wireless things (e.g. connect-disconnect on my Nokia 8), so may be in need of further tuning of those settings, like beam-forming; there are plenty of advises on this subject around here)

My next step is to reinstall Diversion to see whether the problem re-emerges. In that case one may conclude that ad-blocking is somehow responsible for the problem in which case the blacklist may be too stringent, as @ColinTaylor assumed.
If that doesn't happen I will try the ad-blocking in UnBound once more.

If Diversion doesn't change the currently available stable connectivity to Google that brings back the question why I'm the only one reporting this issue, but let's take it one step at a time.

 

[Rik Stigter]

Secondary test-results:

1. The problem is back with ad-blocking with Diversion. Will keep that but...
   
2. Uninstalled UnBound and reverted to standard DNS settings, i.e. ISP-DNS (cf. @L&LD) manually set, no DNS-filter, no DNSSEC and no DoT.

Now wait...

 

[Rik Stigter]

Tertiary test-results:

1. With Diversion installed Google gets really slow after the (in-)famous 2-4 hour period. Uninstalled Diversion
2. Went back to basic config, so removed Skynet as well

The only thing left is the Swap-file, all other amtm-adaptations are now gone. Will keep this running for a while. After reboot everything is up to speed once again. But for how long?

PS. I do check log-files but until now they haven't told me anything useful.