[R7800] problems since fiber

[HELLO_wORLD]

Hello everybody,

I just switched from ADSL to fiber (symmetrical 1 Gb).
The router is connected directly to the ONT, and WAN setup connection is made from DHCP.

It works somehow, because I do get connectivity, my static IP address, gateway, subnet and DNS are given.

The problem is that the router/connection is instable.
When I ssh to it, it randomly freeze (connection lost). htop does not show crazy CPU usage, and nothing special is going on in the LAN network. I often see netconn.sh being called several times.
However, the router does not crash.
The log shows unusual WAN reconnection:

[Internet connected] IP address: 2.xx.xx.xx, Wednesday, July 29, 2020 12:41:56
[Internet connected] IP address: 2.xx.xx.xx, Wednesday, July 29, 2020 12:39:26
[DHCP IP: 192.168.0.3] to MAC address xx:xx:xx:xx:xx:xx, Wednesday, July 29, 2020 12:39:23
[DHCP IP: 192.168.0.3] to MAC address xx:xx:xx:xx:xx:xx, Wednesday, July 29, 2020 12:39:22
[DHCP IP: 192.168.0.39] to MAC address xx:xx:xx:xx:xx:xx, Wednesday, July 29, 2020 12:38:55
[Internet connected] IP address: 2.xx.xx.xx, Wednesday, July 29, 2020 12:36:55
[Internet connected] IP address: 2.xx.xx.xx, Wednesday, July 29, 2020 12:31:55
[Internet connected] IP address: 2.xx.xx.xx, Wednesday, July 29, 2020 12:29:24
[Internet connected] IP address: 2.xx.xx.xx, Wednesday, July 29, 2020 12:26:55
[Internet connected] IP address: 2.xx.xx.xx, Wednesday, July 29, 2020 12:19:24
...

2.xx.xx.xx is my static IP address given by ISP.

Using latest @Voxel firmware.
I had no such problem with previous ADSL (bandwidth was about 20 Mbit/s down and 1 Mbit/s up, now it is 1 Gbit/ up and down).
WAN setup was different: connection was manual static IP and gateway to ISP modem (LAN side of it). Now, WAN connection is DHCP direct to ONT (fiber to Ethernet converter).
Pretty much all options are off on router (no upnp, no traffic meter, no QOS, etc...) like it was before.

Any idea on what is going on? Problem with router or on ISP side? Before involving the ISP, I want to make sure it is not coming from my side.

 

[Zolty]

I wonder if it is R7800 problem or ONT problem.
Can you check what happens when you set WAN connection manualy (IP, gateway, subnet)?
The way to check if this is ONT, eould be connecting PC/laptop directly to ONT and monitor what happens (like pinging ONT.for 10 minutes).

 

[HELLO_wORLD]

Ok, it appears that the IO (infrastructure operator, owner of the fiber network and the hardware directly at the other side of my fiber) is relaying the DHCP from the ISP, but if ISP lease is 12 hours, the IO is making 5 minutes leases.

When router starts it fires udhcpc like that (normal):
udhcpc -b -i brwan -h /tmp/dhcp_name.conf -r 0.0.0.0 -N XX.XX.XX.XX
XX.XX.XX.XX is my WAN fixed IP address.

Each time the DHCP client on WAN side is renewed, it seems to reset the DHCP server on my LAN side as well (hence disconnecting my ssh connection to router).

1) Is there a way to force LAN DHCP server not to reset when WAN DHCP is renewed?
2) Is there a way to force WAN DHCP lease duration ? I tried changing in nvram dhcpc_lease_time=300 to something else, but it comes back to 300.
I would like to try longer duration first, and also shorter (150).

 

[HELLO_wORLD]

I wonder if it is R7800 problem or ONT problem.
Can you check what happens when you set WAN connection manualy (IP, gateway, subnet)?
The way to check if this is ONT, eould be connecting PC/laptop directly to ONT and monitor what happens (like pinging ONT.for 10 minutes).

Cannot set WAN manually, it has to be DHCP to have internet. ISP confirmed that (even with same IP, gateway & subnet).

I will try connect directly to ONT, but don’t have a computer around with Ethernet port... Have a raspi somewhere, but I need to set it up for that. And I cannot mess with internet too long.

 

[kamoj]

I had to add this code to my add-on to enable lease renewal for sky iptv.
You need to let it through the firewall.
Maybe you can get some ideas from that:

iptables -I INPUT -p udp --sport 67 --dport 68 -j ACCEPT

My nvram settings:

nvram show | grep lease
lan_lease=86400
wan_lease=86400
dhcpc_lease_time=10800
dhcpc_lease_obtain=18944

 

[HELLO_wORLD]

My dhcpc_lease_time is always going to 300 at each renewal.
The router log shows renewal period is not predictable. Most of the time it is every 150 seconds (expected behavior 1/2 the lease), but sometimes it is 3 minutes, 5 minutes, 7 or even 20 minutes. No reasons.
Why this inconsistency?
Since my WAN IP, the WAN gateway and the subnet are always the same, I would like to trick the ONT.
DHCP on WAN side is mandatory, without it, no internet even with right IP/gateway/subnet set.
I would like to set the router on static manual configuration but keep the DHCP client (udhcpc) running on the router (WAN siand force asking for leases every 150 seconds (without changing the actual real settings of the router, just sending it’s requests to ONT every 150 minutes).


DHCP is mandatory because it keeps my IP authorized on the ISP side (more exactly the fiber optics network operator who is on the other side of my fiber on behalf of my ISP).

Is that possible?

 

[kamoj]

You can try this and see what happens: infinity lease
You can make the age forward/backward by manipulating dhcpc_lease_obtain.

nvram set dhcpc_lease_time="infinity lease"

nvram set dhcpc_lease_obtain="$(cat /proc/uptime | awk -F. '{print $1}')"
# Or fake obtain time:
nvram set dhcpc_lease_obtain="$(($(cat /proc/uptime | awk -F. '{print $1}') - 1000))"
nvram set dhcpc_lease_obtain="$(($(cat /proc/uptime | awk -F. '{print $1}') +1000))"

nvram commit

 

[HELLO_wORLD]

For now, I changed settings on the DHCP server on LAN side.
Changed auto_time from 5 to 7200 and it is more stable on LAN side, I don’t lose randomly connexion with router ssh anymore. Apparently, each time the DHCP was renewed on WAN side, it would trigger renewal on LAN side as well. So far so good.

I will dig into tweaking DHCP WAN side a bit later, I am too busy lately (hence less present on this forum, but I am thinking of you all often). I have improved aegis but still have to do some work on it before releasing new version.

 

[HELLO_wORLD]

Ok, udhcpc in firmware /sbin is messing with LAN side as well, and when launched it cuts my ssh connexion to router.
However, no such problem with optware udhcpc (and more recent)

so here is what I did: set router on static config (ip, gateway, subnet)
Created 2 cron jobs:
*/2 * * * * /opt/sbin/arping -c 2 -I brwan GATEWAY_IP
*/2 * * * * /opt/sbin/udhcpc -q -i brwan -r STATIC_IP
One to maintain ARP credentials
The other to maintain a DHCP lease with ISP for my ip.

It works well so far.

 

[HELLO_wORLD]

Ok, now I have a question regarding router management of symmetrical 1 Gbit/s connection.
Doing a speed test from the router, using best server, I easily get the max speed for upload (950 Mbit/s).
However, for download, I get around 720 Mbit/s (and only 650 If aegis is on ). CPUs are at 100% so I wonder if something on the router is limiting the test. ISP connexion in my area should reach 950 Mbit/s for download as it is for upload.

Anyone able to reach 900+ Mbit/s download with this router on a gigabit/s internet connection?
I have very little time to mess around and play with the router lately, and I can’t try stock vs Voxel for example.
Anyone having similar experience?
I have minimal things running on the router except stubby (but disabling it does not improve speed).

Any idea?

 

[HELLO_wORLD]

Today, I had to unplug the router and the ONT for a little while to rewire the cables neatly.
When I replugged and restarted everything, I got 950 Mbit/s up and down.
I turned back aegis on and it does not affect speed (still 950 Mbit/s).

I have no idea what was happening before (and I had restarted the router), and why the download speed was so impacted (720 Mbit/s only with nothing running and 650 with aegis running). It made no sense.

Now, everything is at it should, and I am glad to see that aegis is not impacting traffic speed.

I suppose something was slowing down the router or iptables rules, but htop did not show anything going on.
I will keep an eye to see if the problem will come back or not.

Have a great day, all of you

 

[HELLO_wORLD]

Update:
I was able to reproduce the problem. It is very strange.

If I use any Cat 6 Ethernet wire between ONT and R7800, I got:
- 950+ Mbit/s upload speed
- 720~ Mbit/s download speed (650~ with aegis on)

If I use an old Cat 5 Ethernet wire between ONT and R7800, I have:
- 945~ Mbit/s upload speed
- 945~ Mbit/s download speed (with or without aegis)

So it seem to good quality of a cable is maxing upload speed, and is jamming the router for download speeds (and makes them very dependant on iptables rules complexity).

Anyone having an explanation for this? And having the same experienc?

 

[L&LD]

The Cat5 cable is of better quality/construction and/or the ends/termination are more compatible with the connected equipment than the Cat6 cable(s) you're using.

If you left the ONT and router unplugged from the power for over 10 minutes, that helped too (by allowing the connection to the ISP to be brought up to current standards, from their end).

 

[HELLO_wORLD]

My mistake...
Cat 5 cable was bad... and sync between ONT and router was 100 base T
Did not notice because WAN led does not change to orange like LAN ones.
Speed test was returning 5 digits instead of 6 and I did not realize... So it was really 94 Mbit/s

back to Cat 6, and have 954 Mbit/s up and 740 Mbit/s down (aegis off).

Will test Cat7.
Also will need to do a speed test outside of router to see if it is R7800 that is not able to handle full gigabit download or the connexion that is not fully gigabit.

 

[KW.]

ah get a r9000 so can you fix suricata I need your expertise there.