[Release] Asuswrt-Merlin 384.15 (and 384.13_4) are now available

[RMerlin]

I keep having the mysterious client isolation on ASUSWRT-Merlin RT-AC3200 384.13_4

Try disabling Smart Connect. Then, check all three bands to make sure that isolation is isolated on all three - chances are that it's not disabled on one of them. After that, re-enable Smart Connect.

 

[LegitPC]

Upgraded an AC86u from 384.13 to 384.15 via teamviwer and the router never turned back on.

I'll probably be on-site tomorrow morning to fix it, anything I should report back to prevent issues for others?

I've advised the site to turn off the router and unplug a usb drive with jffs scripts like amtm on it and turn it back on.

 

[Val D.]

I'll probably be on-site tomorrow morning to fix it, anything I should report back to prevent issues for others?

Do not reboot this router remotely and do not set scheduled reboot on it, known bug:
https://www.snbforums.com/threads/ac86u-sometimes-powers-completely-off-during-reboot.48296/
Tell the people there just to press the power button off/on to boot the router. Nothing else is usually needed.

 

[CaptnDanLKW]

Try disabling Smart Connect. Then, check all three bands to make sure that isolation is isolated on all three - chances are that it's not disabled on one of them. After that, re-enable Smart Connect.

I see a lot of recommendations to disable SC. I had really good luck for a couple months when on the 384.13 branch, then struggled with 2.4Ghz IOT devices on my isolated Guest network falling off and staying off after going to _14 and _15. Since then I've also put AI Mesh into play.... so I've since disabled it as well.

Is there consensus that SC itself is broken / flawed (i.e. Asus could potentially eventually fix)?

Or is it just that what it does (channel changes, band steering, etc.) causes sensitive devices with poor implementation of cheap WiFi chipsets to freak out? If is didn't have a bunch of smart lights I'd never have noticed a problem with SC.

 

[Mrk26]

Unfortunately that how smart connect working, playing up with chanels and frequencies to avoid congested channels and frequencies. In busy areas it will work harder causing disconnections more often, it will jump on frequencies that are not congested but some devices don't support (especially old) all of them.

 

[CaptnDanLKW]

Unfortunately that how smart connect working, playing up with chanels and frequencies to avoid congested channels and frequencies. In busy areas it will work harder causing disconnections more often, it will jump on frequencies that are not congested but some devices don't support (especially old) all of them.

Understood about Auto Channel selection, however every router does that. When any brand router changes channels the client will move over to it without disconnecting and re-association, assuming its not total garbage. (Right?)

However Smart Connect does more (but auto channel selection is one of them), so I'll re-phrase - is there something wrong with SC, or it is just the sole fault of poor wifi software stacks in some clients?

 

[Asad Ali]

is there something wrong with SC, or it is just the sole fault of poor wifi software stacks in some clients?

Nothing is wrong with SC and I'm using it without any issues for the last 3+ months. You'll just have to experiment with the settings to find the perfect recipe for your environment.

 

[podkaracz]

Nothing is wrong with SC and I'm using it without any issues for the last 3+ months. You'll just have to experiment with the settings to find the perfect recipe for your environment.

You are contradicting yourself , if everything works as expected why should I experiment to get it working decently?

 

[CaptnDanLKW]

Nothing is wrong with SC and I'm using it without any issues for the last 3+ months. You'll just have to experiment with the settings to find the perfect recipe for your environment.

My OP actually mentioned I had it working with no issues for a couple months, so I had it stable. And with SC its enabled or not - if enabled theres nothing to tweak.

I don't want to blame the change on the firmware upgrade since the Wifi driver from _13 -> _15 has not changed. Did a complete reset and manual reconfigure when I went to ._14.2 as a test.

I live in a very light density area and can only see 1 or 2 other reachable networks so I shouldn't have a lot of channel changes anyways.

I've read dozens of threads and hundreds of replies with so many recommendations that work for them but it doesn't get to the crux of my question. Guess it will be one of life's greatest mysteries.

 

[Asad Ali]

You are contradicting yourself , if everything works as expected why should I experiment to get it working decently?

Because it requires proper setting, SC is not a magical one click all set button, you have to properly calibrate it for your network.

 

[Asad Ali]

And with SC its enabled or not - if enabled theres nothing to tweak.

Who told you that?
Asus gave quite a large number of fine tuning knobs for SC, they're under Network Tools > Smart Connect Rule. ( poor placement of the rules page, I know lol )

 

[RMerlin]

Or is it just that what it does (channel changes, band steering, etc.) causes sensitive devices with poor implementation of cheap WiFi chipsets to freak out? If is didn't have a bunch of smart lights I'd never have noticed a problem with SC.

The idea is just to temporarily disable it so you get access to all separate Wifi settings, make sure Isolation is disabled for all three bands, and then re-enable Smart Connect, and see if it works any better.

 

[unsynaps]

Upgraded an AC86u from 384.13 to 384.15 via teamviwer and the router never turned back on.

I'll probably be on-site tomorrow morning to fix it, anything I should report back to prevent issues for others?

I've advised the site to turn off the router and unplug a usb drive with jffs scripts like amtm on it and turn it back on.

That router has a know issue when soft rebooted (through UI or SSH) to not power back on sometimes.

Just cycle the power button on the back left or unplug it and plug it back in.

 

[CaptnDanLKW]

The idea is just to temporarily disable it so you get access to all separate Wifi settings, make sure Isolation is disabled for all three bands, and then re-enable Smart Connect, and see if it works any better.

My devices are AC86Us so they are only 2 bands (right)? I don't recall an 'Isolation' setting and disabling SC only enables the changing of Mode, Width and Control Channel - no new options appear.
(I should also mention that I disable universal beamforming and airtime fairness for both bands as one of the many consistent recommendations made)

 

[mlippold]

My devices are AC86Us so they are only 2 bands (right)? I don't recall an 'Isolation' setting and disabling SC only enables the changing of Mode, Width and Control Channel - no new options appear.
(I should also mention that I disable universal beamforming and airtime fairness for both bands as one of the many consistent recommendations made)

The "Set AP Isolated" is in the Wireless -> Professional area (with drop downs for each band).

 

[Luizlp10]

Came from 384.15 Beta_1 and just now reached the 7 day milestone with 384.15 on my AC86u. I followed L&D's M&M config at first and then slowly started to add scripts, one at a time. Diversion, FreshJr_QoS, Scribe and uiScribe are running smoothly without any noticeable issues and I set a 2 GB swap file on my USB drive.

Will keep track and provide later information for you guys but so far, so great. Thanks, Merlin.

 

[JGrana]

Unfortunately, count me in the camp of users that needs to downgrade...
My setup and config - AX88U as Wireless Router and AC88U and AC68U as Media Bridges.
AX88U - separate 2.4 and 5Ghz SSIDs, fixed channels (48 for 5Ghz, 9 for 2.4Ghz). 2.4 is on the 20/40 setting, 5Ghz is 80Mhz.
No QoS, running spdMerlin, amtm and scribe. 5 devices hard wired, around 16 clients on 2.4 band and around 18-20 on the 5Ghz. 2 of the 5Ghz devices are AX200 based (so 16 AC and 2 AX). My ISP (Spectrum cable) provides 100 down, 10 up via an Arris SB6183 cable modem.

What the issue was is a gradual decay of upload speed to the point it almost stops. Download appears to be fine. Its so bad, even DNS requests can fail. Here is a graph from spdMerlin showing a typical 24 hour period running 384.15:



After numerous tests, the only way to get back to "normal" is a reboot of the AX88U. I have done the factory reset and restored my settings by hand - no change.

After ~a week of this, I re-installed 384.14, did the factory reset and settings. It has been approx 24 hours and it is running much better. Here is a spdMerlin graph running 384.14



All I can suspect at this point is the newer Asus code merge in 384.15 for the AX88U.
I noticed 384.16 Alpha is out - and from Merlins comments will be using another Asus merge. I will try that once the new Asus code is merged...

I really liked 384.15. The promise of OFDMA along with better amtm/Add-ons was great!

 

[dave14305]

Regarding the addon API, in the am_get_webui_page function, the md5sum will never match if the return URL is modified in the destination directory per the wiki instructions.

# Set correct return URL within your copied page
sed -i "s/MyPage.asp/$am_webui_page/g" /www/user/$am_webui_page

Peeking at @Adamm code and @Jack Yaz code, they do not set the current or next page, so it’s not an issue. I’m not smart enough to second guess Merlin’s template. What are the pros and cons?

 

[RMerlin]

Regarding the addon API, in the am_get_webui_page function, the md5sum will never match if the return URL is modified in the destination directory per the wiki instructions.

# Set correct return URL within your copied page
sed -i "s/MyPage.asp/$am_webui_page/g" /www/user/$am_webui_page

Peeking at @Adamm code and @Jack Yaz code, they do not set the current or next page, so it’s not an issue. I’m not smart enough to second guess Merlin’s template. What are the pros and cons?

That would indeed make the md5 check useless if the page is properly implemented (i.e. it gets the correct current_page and next_page variable set). It probably works for the devs currently using it because they don't fully allow applying settings despite not setting these two variables properly. I guess I should remove that parameter from the helper script.

 

[RMerlin]

AX88U - separate 2.4 and 5Ghz SSIDs, fixed channels (48 for 5Ghz, 9 for 2.4Ghz). 2.4 is on the 20/40 setting, 5Ghz is 80Mhz.

Two problems right off the bat:

1) 48 is a DFS channel, which means at any time the clients might get bumped off it if a radar is detected
2) Don't use 20/40, as most people will have too much interference for 40 MHz to work, and will end up downgrading to 20 MHz anyway