380.57 Possible port forwarding issue

[Soul--Reaver]

Ok then. This is the steps I will take.

1. Reset
2. Update FW
3. Reset
4. Add ports and stuff manually
5. Disable NAT Acceleration
6. Reboot
7. Test if problem persists.

I'll post the output here.

Daniel i will be testing the NAT loopback setting tonight. Would you be willing to add that as well to the above steps?

 

[RMerlin]

Thanks! Will try this later, maybe tomorrow. Downside is that sites will be slower on my server. Will this be fixed in upcoming versions?

It won't make any difference to your visitors. All port-forwarded traffic bypasses CTF already as it's not compatible, so they are never accelerated. Only your regular NAT traffic is.

 

[Daniel Hansson]

Ok, seems like disabling NAT Acceleration did the trick for me. What I don't like is that traffic is slower now. I have around 7-900 visitors per day, and when I did new tests the loading time increased by 10 seconds approximately without NAT Acceleration (CPU is overwhelmed) - so a fix in the next release would be appreciated!

This also affects my upload speed which is 120 MB/s normally, and now it's down to ~60 MB/s

 

[Soul--Reaver]

Alright i completely reset the router to factory

When i did just the basics to set up the webserver port forward everything worked

When i finished setting up the rest of the configuration the port forwarding was broken again

I will go through this setting by setting and test it every time. See if i can find the thing that breaks it

 

[Soul--Reaver]

Daniel how many port forwards have you set up? when i only add the webserver everything works it seems

When i add others it seems to break

 

[CiscoX]

Which port are you trying to open? Did you also try the NAT loopback setting from Merlin to ASUS or ASUS to Merlin.

 

[Soul--Reaver]

TCP port 16000

Still testing the port forwards. Will also test NAT loopback

 

[Soul--Reaver]

Alright it seems that the more port forwards i add the slower the connection setup process becomes. Actual transfer speeds stay good

Currently i have one second delays on new external connections and the port forwards look like this. I still need to add 5 more

Virtual Servers
Destination - Proto - Port range - Redirect to - Local Port - Chain
ALL - TCP - 80 - 192.168.1.100 - 80 - VSERVER
ALL - TCP - 21 - 192.168.1.101 - 21 - VSERVER
ALL - TCP - 51000:51050 - 192.168.1.101 - 51000:51050 - VSERVER
ALL - TCP - 16000 - 192.168.1.100 - 16000 - VSERVER
ALL - TCP - 14000 - 192.168.1.100 - 14000 - VSERVER
ALL - UDP - 14500 - 192.168.1.100 - 14500 - VSERVER
ALL - TCP - 14801 - 192.168.1.100 - 14801 - VSERVER
ALL - UDP - 14801 - 192.168.1.100 - 14801 - VSERVER

 

[Soul--Reaver]

NAT Loopback has zero effect on the situation... at least for me

Tried both options

 

[Daniel Hansson]

Ok, so after some more investigation I decided to revert back to 378.56_2, the old FW. Reasons why is that when I left it on for like 2 hours I got the same effect as with the new FW. The reason for the same effect is (what I think) that the CPU can't handle the traffic and the continuous downloads from my server + the traffic from all the visitors. I have a ASUS-AC66U, and it's a good router, but the traffic is just to much to handle without NAT Acceleration so the effect became that the site was offline as before - sometimes it worked, and some not, but mostly not.

With the old FW I have no issues at all, and since I now tested several times (did some tests before as well) my decision is to use the old FW. Please fix this irritating bug in the next release! Thank you.

 

[Soul--Reaver]

Daniel how many port forwards do you have?

 

[Daniel Hansson]

Daniel how many port forwards do you have?

I tested with 3 - 80, 443, and 22. But I use 6 forwards with the old FW.

 

[Soul--Reaver]

Well with 3 or 4 port forwards i could not find any issues in short term tests. With 7 or 8 i noticed immediate problems

 

[Daniel Hansson]

As I stated earlier - I think the router can't handle the bandwidth without the NAT Acceleration in my case, and with NAT Acceleration it fails due to the original issue with the new FW. I don't know if you have that saying, but in Sweden we call it "Moment 22 - Catch 22".

Anyhow, the issue should be addressed.

 

[Soul--Reaver]

All issues should be fixed. But they cannot be fixed as long as you don't know the cause. Reproduciblity helps a lot with finding that.

Still remains that its likely something also present in the original firmware which would mean that Asus or related partners should fix it as RMerlin probably can't
Note to self: Need to check Asus firmware

 

[Daniel Hansson]

I totally agree, and understand. I really hope someone nails it so that the issue is fixed.

 

[RMerlin]

Asking me four times to "fix it for the next release" isn't going to achieve anything, as the NAT acceleration code is closed source, and only Broadcom has its source code.

It will be fixed whenever Broadcom fixes it, and passes on any fix to Asus.

 

[Daniel Hansson]

Is it possible to report upstream?

And that is /if/ *that* is the issue here. Likely, but I'm not convinced.

 

[kvic]

Is it possible to report upstream?

And that is /if/ *that* is the issue here. Likely, but I'm not convinced.

Report CTF issue upstream? If you can reproduce in the stock firmware, then can go directly to Asus. Otherwise, it's difficult to build a case.

I don't have AC66 but I'm surprised to hear with CTF on/off, performance to your web server is affected so much..you sure nothing else at play with your setup?

 

[Daniel Hansson]

Report CTF issue upstream? If you can reproduce in the stock firmware, then can go directly to Asus. Otherwise, it's difficult to build a case.

I don't have AC66 but I'm surprised to hear with CTF on/off, performance to your web server is affected so much..you sure nothing else at play with your setup?

Yes. I'm sure. It's superfast normally.