Strange WAN slowdown / breakdown problem with Archer C7 v2/v4

fatzgenfatz · December 8, 2020, 3:13pm

Hi, I'm using my TP-Link Archer C7 v2 as main router for my network. On WAN I have a Gbit-connection to the fiber wire converter in the basement (150/20 internet package).

All my routers run on Openwrt.

To connect another room to my network I use an Archer C50 as WDS-Client which works fine.
Because the C50 only has 100Mbit LAN i bought an used Archer C7 v4 to get GBit-LAN.

Sometimes my WAN connection gets slow (90Mbit) or slower (9Mbit) or disconnects at all. When I unplug and plug in the WAN-Cable everything is normal again (restarting the WAN-connection in Luci doesn't work). I don't know if that problem is caused by my Archer C7 or the Ethernet2Fiberwire box in the Basement (Mikrotik, maintained by my provider).

So I wanted to test if that problem also exists with the Archer C7 v4.
But when I make a backup of my C7v2 and install it on my C7v4 all settings seem to be ok but LAN/WAN does not work at all. I can't get any connection via ethernet.
When I connect via WLAN all my settings look ok, in Luci on the Network/Switch page the connected cables/devices show as "1000baseT full-duplex", but I can't even ping any device.

I tried installing the original TP-Link firmware and everything works fine with that one.

Any ideas what my problem could be?

Thanks in advance,

f.

BIGFAT · December 8, 2020, 5:07pm

You cant use backups files for one router for an entire another version of the same device: v2 ≠ v4 in matter of hardware. You need to reconfigure manually and use the unziped backups as reference. Big jumps in Openwrt releases or big changes in certain packages tend to result in issues too.

Beside that, depending on what you are running on the archer c7v2 or v4 itself (sqm or qos?), i wouldnt except more than stable 90-100mbit on wan. The cpu on them is quite slow.

I would rather focus rebuilding the entire network. Reset everything and use the latest stable. Use the archer c7v4 as a main router, swap from wds to mesh to get better performance over wlan (beside wiki, here is a reference in matter of mesh: Setting up authenticated mesh with wpad-mesh).

If the c7v4 router still crap randomly out on wan or lan, then you will have to check the rest like cabling and any provider device. If the v2 itself dies out as a mesh node, well its time for the bin. I had quite a bunch of them for an hotspot project and they all died out after a few years with similar issues (wan, lan and wlan getting unstable).

Do you have any info on the mikrotik? If its an active device, mikrotik tend to have breaking bugs in the past, and since isps arent trustable to update their own damn equipment...

mk24 · December 8, 2020, 5:57pm

Yes what @BIGFAT said the v2 and v4 are sufficiently different you'll have problems copying config files over expecting them to work the same. Reconfigure from scratch often helps.

Also you could run a single router with stock firmware for a while to see if the Internet connection is stable. Running stock firmware is the standard way to rule out hardware problems.

BIGFAT · December 8, 2020, 6:42pm

Indeed, then another stable or snapshot should be worth a try before going into debugging.

fatzgenfatz · December 8, 2020, 6:59pm

Ok, I will try to configure the v4 from scratch when I find the time for it.
I use the imagebuilder to create nightly images for my routers.
With sqm I usually get 140Mbit which is ok for me.
I never had performance problems with WDS, but I will try mesh (again when I find time for it).
The Mikrotik is a RBFTC11 which was installed about a month ago, I just got the fiber net.

Thanks for your help!
f.

BIGFAT · December 9, 2020, 12:14am

OpenWRT recently jumped from kernel 5.4 to 5.8 on snapshot, and the imagebuilder set on "master"
build on the same principle. It could be it. Try to build on the latest stable (with the needed wpad-mesh-openssl modification).

Probably the absolut max with the recent performance gain from the 5.x kernel. Additional heavy lan or wlan traffic wouldnt be ideal. But again its not the issue here and its something for another topic.

I can only recommend it, it doubled my bandwidth between the nodes after making the change myself last year.

I dont see any problem here, until it got somehow broken. If its installed outdoor you could alway check the installation.

fatzgenfatz · December 9, 2020, 3:16pm

I found the problem: The v2 has eth0 and eth1, the v4 only has eth0. So the config files don't work.
At the moment I'm trying to build an 19.07.4 image with imagebuilder. But I'm a little confused because the kernel is 4.14. The kernel of the stable prebuilt image also is 4.14.

Something seems wrong here
I downloaded https://downloads.openwrt.org/releases/19.07.4/targets/ath79/generic/openwrt-19.07.4-ath79-generic-tplink_archer-c7-v4-squashfs-sysupgrade.bin ->Kernel 4.14
For imagebuilder I downloaded https://downloads.openwrt.org/releases/19.07.4/targets/ath79/generic/openwrt-imagebuilder-19.07.4-ath79-generic.Linux-x86_64.tar.xz -> Kernel 4.14

The snapshot-package I'm using is https://downloads.openwrt.org/snapshots/targets/ath79/generic/openwrt-imagebuilder-ath79-generic.Linux-x86_64.tar.xz -> Kernel 5.4.18

Maybe I downloaded the wrong 19.07.4-Packages?

mk24 · December 9, 2020, 4:10pm

Of course, you're going back to an older version including an older kernel. Snapshots really aren't recommended for "production" deployment around the house unless there is something you need missing or known to be broken in the latest release, or a release doesn't exist for your model.

Image Builder uses the pre-built kernel of its version, the exact same kernel you'd run with a release install. In other words it doesn't rebuild the kernel, it only adds or removes packages from the file system, including kmods.

fatzgenfatz · December 9, 2020, 4:27pm

Ok, I understand. I just thought Bigfat said that snapshot should use 5.8.

BIGFAT · December 9, 2020, 11:56pm

I didnt really check on ath79 tree, or specially your devices about kernel 5.8, sorry for that. But well if it works now for you after backing new images and new configs

fatzgenfatz · December 10, 2020, 9:07am

It works now like before, but I can't chose the mode "mesh" for my wlan. Package wpad-mesh-openssl is installed. I assume this is because kernel 5.4?

BIGFAT · December 14, 2020, 11:50am

Sry for the late response, im curently sick. No its not because of the kernel. Im myself on 5.4 on a similar ath10k wlan plattform. First check if you have the ct version of the ath10k firmware installed on both archers. What you should look after for mesh in luci is called "802.11s", and of cource use wpa3-sae. If you cant find anything, ignore luci and try to make the config yourself and take the link i posted as a reference.

fatzgenfatz · December 14, 2020, 3:00pm

No problem, get well soon!
I tried to create a mesh wlan, but I didn't succeed.
Here is my wireless config:

config wifi-device 'radio0'
        option type 'mac80211'
        option channel '36'
        option hwmode '11a'
        option path 'pci0000:00/0000:00:00.0'
        option htmode 'VHT80'

config wifi-iface 'mesh'
        option device 'radio0'
        option network 'lan'
        option mode 'mesh'
        option mesh_id 'yourmeshid'
        option mesh_fwding '1'
        option key 'secretkey'
        option mesh_rssi_threshold '0'
        option encryption 'sae'

config wifi-device 'radio1'
        option type 'mac80211'
        option channel '11'
        option hwmode '11g'
        option path 'platform/ahb/18100000.wmac'
        option htmode 'HT20'
        option disabled '1'

But Luci says " Wireless is not associated"

BIGFAT · December 15, 2020, 12:02am

Thanks. Choose a country code and tx power fitting your location, device capability and local regulation for both sides. Also change mesh id and pw to something more personal.

Ct firmware and wpad-mesh are there, After those minor changes it should boot right up.

fatzgenfatz · December 15, 2020, 11:06am

Sorry, I don't know if I'm just too dumb?
On my main router the wireless file looks like this:

config wifi-device 'radio0'
        option type 'mac80211'
        option hwmode '11a'
        option path 'pci0000:00/0000:00:00.0'
        option htmode 'VHT80'
        option disabled '0'
        option country 'AT'
        option channel '36'
        option cell_density '0'
        option txpower '23'

config wifi-iface 'mesh'
        option device 'radio0'
        option network 'lan'
        option mode 'mesh'
        option mesh_id 'dasistmeinmesh'
        option mesh_fwding '1'
        option key 'xxxxx'
        option mesh_rssi_threshold '0'
        option encryption 'sae'

config wifi-iface 'default_radio0'
        option device 'radio0'
        option network 'lan'
        option mode 'ap'
        option ssid 'OpenWrt'
        option key 'xxxx'
        option wpa_disable_eapol_key_retries '1'
        option wds '1'
        option encryption 'sae-mixed'

config wifi-device 'radio1'
        option type 'mac80211'
        option hwmode '11g'
        option path 'platform/ahb/18100000.wmac'
        option htmode 'HT20'
        option channel '6'
        option country 'AT'
        option disabled '0'

config wifi-iface 'default_radio1'
        option device 'radio1'
        option network 'lan'
        option mode 'ap'
        option ssid 'Abraham Linksys'
        option key 'xxxx'
        option wpa_disable_eapol_key_retries '1'
        option encryption 'sae-mixed'

config wifi-iface 'wifinet2'
        option device 'radio1'
        option mode 'ap'
        option isolate '1'
        option key 'xxxxx'
        option wpa_disable_eapol_key_retries '1'
        option network 'guest'
        option ssid 'alguest'
        option encryption 'sae-mixed'

But I still get "Wireless is not associated" in Luci.

If I use this as wireless file, Luci won't even show me the wireless page:

config wifi-device 'radio0'
        option type 'mac80211'
        option hwmode '11a'
        option path 'pci0000:00/0000:00:00.0'
        option htmode 'VHT80'
        option country 'AT'
        option txpower '23'
        option channel '36'

config wifi-iface 'mesh'
        option device 'radio0'
        option network 'lan'
        option mode 'mesh'
        option mesh_id 'dasistmeinmesh'
        option mesh_fwding '1'
        option key 'xxxxxx'
        option mesh_rssi_threshold '0'
        option encryption 'sae'

config wifi-device 'radio1'
        option type 'mac80211'
        option hwmode '11g'
        option path 'platform/ahb/18100000.wmac'
        option htmode 'HT40'
        option txpower '20'
        option country AT'
        option legacy_rates '0'
        option channel '13'

On my second router this wireless file is ok with luci, but I still get Wireless is not associated

config wifi-device 'radio0'
        option type 'mac80211'
        option channel '36'
        option hwmode '11a'
        option path 'pci0000:00/0000:00:00.0'
        option htmode 'VHT80'
        option country 'AT'
        option txpower '23'

config wifi-iface 'mesh'
        option device 'radio0'
        option network 'lan'
        option mode 'mesh'
        option mesh_id 'dasistmeinmesh'
        option mesh_fwding '1'
        option key 'xxxxx'
        option mesh_rssi_threshold '0'
        option encryption 'sae'

config wifi-device 'radio1'
        option type 'mac80211'
        option channel '11'
        option hwmode '11g'
        option path 'platform/ahb/18100000.wmac'
        option htmode 'HT20'
        option disabled '1'

Just to understand it better: Do I have to have the mesh-part in both routers?

config wifi-iface 'mesh'
        option device 'radio0'
        option network 'lan'
        option mode 'mesh'
        option mesh_id 'dasistmeinmesh'
        option mesh_fwding '1'
        option key 'xxxxx'
        option mesh_rssi_threshold '0'
        option encryption 'sae'

Or do I have to scan the Wlan on the second router and join the net?

Sorry for my dumb questions

BIGFAT · December 18, 2020, 3:14pm

My last guess: have you removed the other wpad package? You should really only have wpad-mesh-openssl installed. If not: do so and reboot.

If it doesnt work, its time to start a new thread to get more people involved, since your config is clean, i dont see why it doesnt work.

mk24 · December 18, 2020, 4:17pm

I would not trust LuCI to be fully accurate about mesh mode.

On each router, iw dev should show its local mesh point running:

root@batbr1:~# iw dev
phy#0
        Interface mesh0
                ifindex 14
                wdev 0x2
                addr f0:9f:c2:<redacted>
                type mesh point
                channel 153 (5765 MHz), width: 40 MHz, center1: 5755 MHz
                txpower 10.00 dBm
                multicast TXQ:
                        qsz-byt qsz-pkt flows   drops   marks   overlmt hashcol tx-bytes        tx-packets
                        0       0       4278470 0       0       0       0       1406287184              4278626

Mesh point interfaces transmit beacon packets like an AP does, so you can scan to see other mesh point(s) on the air:

root@batbr1:~# iwinfo mesh0 scan
Cell 01 - Address: B4:FB:E4:<redacted>
          ESSID: "<redacted>"
          Mode: Mesh Point  Channel: 153
          Signal: -48 dBm  Quality: 62/70
          Encryption: WPA3 SAE (CCMP)

The channel must match exactly or the mesh points will not connect.

Finally you can see if the mesh point is associated:

root@batbr1:~# iwinfo mesh0 assoclist
B4:FB:E4:<redacted>  -51 dBm / -95 dBm (SNR 44)  10 ms ago
        RX: 300.0 MBit/s, MCS 15, 40MHz             11268852 Pkts.
        TX: 300.0 MBit/s, MCS 15, 40MHz              4141713 Pkts.
        expected throughput: 59.4 MBit/s

Try simplifying the configuration: drop htmode to HT20, remove the AP on the same radio, etc.

BIGFAT · December 18, 2020, 4:36pm

Or VHT20, since he is trying to build a mesh on 5ghz.

fatzgenfatz · December 18, 2020, 5:55pm

Strange, on my Archer C50 mesh seems to work with wpad-mesh-openssl:

root@OpenWrtc50:/etc/config# iw dev
phy#1
        Interface wlan1-1
                ifindex 19
                wdev 0x100000004
                addr da:0d:xxxxxxxxxxx
                type mesh point
                channel 36 (5180 MHz), width: 80 MHz, center1: 5210 MHz
                txpower 22.00 dBm
                multicast TXQ:
                        qsz-byt qsz-pkt flows   drops   marks   overlmt hashcol tx-bytes        tx-packets
                        0       0       907     0       0       0       0       108118          908

But on my Archer C7 v2/v4 it just does not work. I can see the network but not connect to it:

root@OpenWrt:/etc/config# iwinfo wlan0 scan
Cell 01 - Address: DA:0D:xxxxxxxxxxxxxx
          ESSID: "meinmesh"
          Mode: Mesh Point  Channel: 36
          Signal: -29 dBm  Quality: 70/70
          Encryption: WPA3 SAE (CCMP)


phy#0
        Interface wlan0
                ifindex 69
                wdev 0x1a
                addr 50:c7:xxxxxxxxxxxxx
                type mesh point
                txpower 0.00 dBm
                multicast TXQ:
                        qsz-byt qsz-pkt flows   drops   marks   overlmt hashcol tx-bytes        tx-packets
                        0       0       0       0       0       0       0       0               0

On both routers only wpad-mesh-openssl is installed and I rebooted them after changes.

I think I will stick to my working WDS-config since it is fast enough for me. When I have more free time perhaps I will try again.

About my first post in this thread: The speed dropping from 150 to 90 Mbit seems to be a problem on my providers side. I allready changed the cable and at the moment I'm using a Raspberry Pi 4 as router and the Problem exists there too.

Thank you all for your help!
f.