Welcome Guest ( Log In | Register )

Bump Topic Add Reply RSS Feed
6 Pages < 1 2 3 4 > » Bottom

Outline · [ Standard ] · Linear+

 Official TM UniFi High Speed Broadband Thread V43, READ 1ST PAGE FOR RELEVANT WIFI INFO!

views
     
Anime4000
post May 27 2025, 02:24 AM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
It appears Unifi had updated my account VLAN profile, adding VEIP along side usual PPTP, this show in favor of AIO and make old dumb bridge ONU cannot be use if OLT say so

user posted image

By default FwdOp from this OLT is 0x10:
CODE
0x10
Action (h) (VID investigation)
Discarding (c)


What it does with Action H:
QUOTE
h) Bidirectional positive filtering by TCI: Both ingress and egress frames are filtered on investigation of the TCI fields. If and only if the specified fields in the TCI of a candidate ingress frame match an entry in the VLAN filter list, the frame is admitted into the bridge. If and only if the specified fields in the TCI of a candidate egress frame match an entry in the VLAN filter list, the frame is forwarded based on the destination MAC address, according to action a).
Back to Action A:
QUOTE
a) Basic MAC bridge operation: All frames are accepted into the MAC bridging entity.
Egress frames are forwarded from this port if either a) the frame's MAC DA is listed in
the MAC bridge port bridge table data for this port or b) the frame's DA does not appear
in the MAC bridge port bridge table data for any port (flooding). The contents of the
VLAN filter list attribute are not meaningful. See Figure 9.3.11-3.
user posted image

For now, I putting FwdOp 0x02 at VEIP and leave PPTP default, I need update SDK of this

This post has been edited by Anime4000: May 27 2025, 02:24 AM
Anime4000
post May 27 2025, 02:22 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(hsbb @ May 27 2025, 02:14 PM)
Does maxis currently install it own fiber cabinet/pole/DP at ur area?
*
No, just Unifi, all TM speedtest server give me 200Mbps, but I can only achieve 2Gbps download with Google Drive IDM

QUOTE(kwss @ May 27 2025, 10:22 AM)
Chill chill... Nobody is killing dumb ONU.
I think TCI filtering helps prevent a few things:
1. It prevent people with modded ONU from specifying higher priority for their traffic.
2. It prevent people with modded ONU from tagging VLAN they are not supposed to access, eg: sending on Maxis VLAN while being a Unifi subscriber.
*
Well then, I need quickly finishing latest firmware for this issue, make GPON bridge like Managed Switch

Remap VLAN 500 to 621 for example:
user posted image

and it works
user posted image

This post has been edited by Anime4000: May 27 2025, 02:23 PM
Anime4000
post May 27 2025, 08:12 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
Guys, I think my Unifi suffer from LCP terminated by peer, just like my post: 2019 - Streamyx Internet bottleneck, The Legendary LCP terminated by peer
user posted image

It get disconnect when I run constant upload that max out GPON 1100Mbps:
user posted image

Log (tldr)
user posted image

» Click to show Spoiler - click again to hide... «


some people said ibse02.rsh is shit PPPoE AC
Anime4000
post May 27 2025, 10:08 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 27 2025, 08:46 PM)
Did you revert your "Maxis" VLAN?
Can you get the debug output from:
System > Logging
Add Topic:
pppoe
ppp
debug

Where do you upload and for how long?
*
want to make a report to TM, I set D-Link as Bridge Mode, so they can use TR069
all correct VLAN

test since 3PM until today

have facing this issue in 2019 during streamyx time,
so, I redo what I did before

user posted image

before was ibse02.mc
now was ibse02.rsh

my friend said ibse02.rsh is old where streamyx era exist

lol
Anime4000
post May 27 2025, 10:10 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
kwss down again

user posted image
Anime4000
post May 28 2025, 12:47 AM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 27 2025, 11:10 PM)
So today you started connecting to a different BNG and have this problem?
Did the Mikrotik debug log says something like LCP-Echo timeout or other error?
*
pppoe-out1: sent LCP TermAck id=0x4

what do you think?
Anime4000
post May 28 2025, 01:55 AM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(hsbb @ May 28 2025, 01:38 AM)
Anime4000, can try what neekun already did above but with ur dlink MAC. Let see what will happen.
*
didn't work, my speed not slow, just PPPoE terminate my session, been uploading 6TB of zero (uptime: 22hrs):
user posted image

Even with this ONU that I got from tm friend still didn't work
user posted image

user posted image

Last Saturday my friend checked something wrong with PPPoE, that is why I check now,
user posted image
user posted image

Next morning I ask him again this, and I already send email to TM with relevant documents

my other friend can even choose to avoid ibse02.rsh, I ask around seem this been in service since Streamyx?
user posted image

UPDATE May/28/2025 02:11:27
It's happening again
user posted image

This post has been edited by Anime4000: May 28 2025, 02:15 AM
Anime4000
post May 28 2025, 02:55 AM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(hsbb @ May 28 2025, 02:27 AM)
SN_G004/C042/DP0005 .. dlink in combo mode also session terminated?
*
Yes, terminated by Unifi,

Oddly enough when use 1GbE ONU like HG8240H5 that I have during Maxis time, still disconnected,

I sense that what ever PPP echo did not reach to Unifi,
and Unifi assume offline? and request terminate session

never had this issue when before with ibse02.mc

why unifi switch PPPoE AC that even worse?

Perhaps I need to limit:
2.5G Egress to 1100Mbps (match OLT)
PPPoE Upload to 1000Mbps (match Radius)

give PPP echo overhead never missed an echo

I wondering this would solve it?
might need to disable Hardware Acceleration for QoS to work

This post has been edited by Anime4000: May 28 2025, 02:58 AM
Anime4000
post May 28 2025, 03:40 AM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 28 2025, 03:25 AM)
sent LCP TermAck id=0x4
This is your router acknowledging the termination request. The debug log should show the reason for termination when it receive that request. Can you find it?

Since both BNG show up in scan, you can hard code the BNG name as AC Service in pppoe. Mikrotik should be able to connect to it specifically.

From the MAC address, looks like this broken BNG is not using VRRP setup. Perhaps TM misconfigured it? I think it should be part of VRRP.
Did this BNG allow MTU 1500?

Edit:
Since it says lost carrier, could be broken OLT port or line card too. If your friend can swap your to another line card that would immediately rule it out
*
user posted image

user posted image

user posted image

1492 MTU not work, it will use 1480 in disappointment because VRRP
user posted image
Anime4000
post May 28 2025, 11:12 AM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 28 2025, 09:30 AM)
Anime4000
Just looking at the log, it seems TM configured their BNG to kill the connection if it's just missing one LCP-Echo request / response pair. Default configuration for BNG is supposed to be missing 3 pairs before disconnection.

It also looks like there is a bug inside their BNG software. It didn't actually send the LCP-Echo. From your log, the Echo is no where to be found and suddenly boom, you get a termination request.

Since you are just uploading, it shouldn't block your download direction from receiving the packet. The timing for the termination request is the time slot where the echo request should be. Really looks like software bug to me more than configuration error.

I also don't quite understand why during the pppoe scan, it didn't show VRRP address but it did connect to VRRP address that's no where to be found.
*
That's what I think, ibse02.rsh is stupid bug

since T-CONT set to 1100Mbps and PPPoE 1000Mbps, I try do this:
sfp-sfpplus1 Egress: 1100Mbps
simple queue pppoe-out1: 1000Mbps

see what happend

user posted image
Anime4000
post May 28 2025, 11:31 AM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
kwss

Egress:
user posted image

Simple Queue (here goes my Mikrotik CPU, go up 70% when enable)
user posted image

it did limit L3, 1100Mbps T-CONT within 1024Mbps PPPoE
user posted image
Anime4000
post May 28 2025, 02:20 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
kwss hsbb

After I do fq_codel QoS on PPPoE Interface:
user posted image

been 2hrs no disconnect

However... This is wrong answer and wrong solution to compensate ibse02.rsh
Customer CPE are not suppose to solve bug on ibse02.rsh
I redo like before 2019 - Streamyx Internet bottleneck, The Legendary LCP terminated by peer
I keep reporting and involve MCMC which TM unable to solve this, and wrong MTU 1488 value
Anime4000
post May 28 2025, 02:33 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(tng55 @ May 28 2025, 02:26 PM)
daam trouble for TM

QoS enable no good due cpu hike
need disable QoS i know
i hope you still during troubleshot more project
go ahead MCMC report with document about troubleshot PPPoE disconnect
*
Indeed, AI also agree:
user posted image

hsbb
TM need double check with their vendor, also same answer kwss and AI per RFC, PPPoE AC should never terminate one missed LCP
user posted image
Anime4000
post May 28 2025, 03:08 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 28 2025, 02:54 PM)
Anime4000
Did you manage to find out which vendor is that BNG?
The problem can simply be solved by either:
1. Waiting for 3 echo timeout before sending the termination request. Since one or two broken timeslot won't led to termination.
2. Fix the latent bug where the BNG send the wrong request during the time slot. This kind of bug sounds like race condition in those super old C code. The system behave differently with different load. In your case, I suspect HQoS in your profile play a role.
If you QoS on your side and didn't trigger the shaper / policer on the BNG, then it won't mix up the time slot.

Actually it's just software, can request your TM friend to just upgrade it? Still have support contract right?
The VRRP problem fixed itself... Interesting.
*
It's Juniper Network handle the PPPoE here,
if that case, increase waiting echo to 5 times should be suficient, just in case everyone upload in same PON cause LCP packet loss

I try use HG8240H5 ONU, limiting my self to synmetrical 1Gbps (1GbE)
it still happening when I fully saturate the 1GbE port, cause LCP did not reach either way

This post has been edited by Anime4000: May 28 2025, 03:09 PM
Anime4000
post May 28 2025, 08:27 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 28 2025, 06:04 PM)
I look up the BNG mac address c4:69:f0:69:da:c6.
It's actually Huawei. If it's Juniper then you should have no problem doing MTU 1500.

If you can find someone to reconfigure the timeout then that should mitigate the problem.

MAC prefix registered on 17 September 2020. So yours is not leftover junk from streamyx era.
*
that's odd, he told me was F0:1C:2D:22:87:C3
he lied to me?

EDIT:
I believe kwss, it make sense for MTU issue when PPPoE over VRRP and VXLAN.
I gonna make report to MCMC that TM staff lied to me and don't want to solve

This post has been edited by Anime4000: May 28 2025, 08:32 PM
Anime4000
post May 28 2025, 10:35 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 28 2025, 09:04 PM)
What exactly did they tell you is the problem and their proposed solution?
*
I asked what PPPoE-AC brand
They said Juniper
Then I reply, Juniper? (in-disbelieve Juniper make PPPoE 1488 bytes MTU)
Then I ask, what is the MAC Address
They reply "F0:1C:2D:22:87:C3"

I know that same mac address that I use before with Maxis, which can MTU 1500 and never had disconnect

then I assume Juniper follow Huawei to virtualize everything, VRRP, VXLAN without compensate PPPoE 8 bytes overhead

they try ask other team

I haven't make a report to MCMC, I need to really make sure not my Mikrotik is the issue

but when I use Queue Tree on pppoe-out1 limit at 1024Mbps, run BTest for 6 hours, never had disconnect


QUOTE(hsbb @ May 28 2025, 09:36 PM)
Anime4000
When there's 02, meanwhile there's 01 exist as a backup. Maybe u can ask ur friend to reroute ur account to ibse01.rsh if exist. But maybe need ur complain ticket to do that.

To reconfigure ibse02.rsh, i don't think so because maybe can risk whole Seremban unifi users will offline.

* U need to move to Melaka now f u want to use ibse02.mc. I think ibse02.rsh not physically exist during the time u connected to ibse02.mc before.
*
I have asked this, they told me cannot, (face with this crap ibse02.rsh)

I assume possible for OLT PON Port that I reside to use different PPPoE AC? some people run PPPoE Scan and have two:
user posted image
not my picture

I ask around, CPE not suppose to fix, ISP need to validate and stress test make sure LCP timeout within 3 to 5 times before issuing LCP terminate

Just don't configure aggressive LCP timeout
Anime4000
post May 30 2025, 10:26 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
kwss hsbb

today afternoon technician come to my home, expected this my issue need ask Unifi Elite

upon configure D-Link as normal (GPON Router), PPPoE doesn't up, thus cannot test
I expect something with VLAN 500 on VEIP

technician planned to give new 2.5 skyworth for free and no need take D-Link back but no stock that time

what he told me:
1. D-Link ONR A1 and B1 is EoL on TM eyes
2. D-Link ONR not play well with multi vendor OLT
3. OLT and PPPoE-AC QoS optimize for Skyworth 2.5G,
4. That optimization cause D-Link become broken by a lot!

I need call 100 this Sunday for Unifi Elite and get Skyworth 2.5 (now toy to reverse engineer)
Anime4000
post May 30 2025, 10:37 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(blacktubi @ May 30 2025, 10:30 PM)
This QoS is causing a lot of issues with other routers especially on upload speed
*
Indeed! many victim of D-Link now using PON Stick

ZTE, Nokia can accept 16 T-CONT that I spam and upload speed is fixed

when I do same with Huawei OLT, O5 without VLAN given, so...
I had to reduce T-CONT to 4 in disappointment

This post has been edited by Anime4000: May 30 2025, 10:37 PM
Anime4000
post May 30 2025, 11:38 PM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 30 2025, 11:12 PM)
My point of view is this: GPON is never designed for anything more than 1Gbps.
Imagine how over-subscribed is each OLT port.
They just keep using dirty hack to make it work.

None of the big telco equipment supplier has GPON with speed faster than 1Gbps. All these smaller player just do custom build for TM.

How is it possible that Skyworth can "support" whatever QoS they meant but D-Link cannot? They use exactly the same SoC.

What I think happened is during D-Link contract, they haven't deploy these kind of dirty hack. After X number of unit is delivered, the contract is done, hence EOL. TM never have support contract with them to begin with, it's just PO for X units. All the D-Link firmware update is just bug fix, no new feature, no support of whatever QoS nonsense TM is doing
*
Indeed, D-Link DPN-FX3060V now use as is, no more update I assume after this, and security update never a thing, even Asus fall into ViciousTrap

QUOTE(hsbb @ May 30 2025, 11:02 PM)
U need to ask later what does it mean by optimized. Also multi vendor OLT shouldn't be optimized for 1 vendor only i.e Skyworth.
*
I beg to differ:

D-Link in Bridge Mode to RB5009 and Acer Predator Helios Neo (Killer 2.5G)
user posted image

Skyworth GN360V (Router Mode) to Acer Predator Helios Neo (Killer 2.5G)
user posted image

due to low speed on D-Link, I try download a file from Google Drive:
user posted image

Now I confused... rclxub.gif speedtest fail, but Download via IDM and upload via BTest correct results

due to I can't use D-Link as Router, as PPPoE never up, FTC mess up my VLAN, now I have VoIP 400 to LAN 4 with Tag and Untagged 😅:
user posted image

user posted image

since I have Skyworth by someone, I still need new Skyworth from TM this monday

This post has been edited by Anime4000: May 30 2025, 11:43 PM
Anime4000
post May 31 2025, 12:29 AM

Look at all my stars!!
*******
Senior Member
2,399 posts

Joined: Jul 2009
From: /dev/null
QUOTE(kwss @ May 31 2025, 12:02 AM)
I have a theory for why this happens.
In this whole world, only Nokia ASIC can do HQoS. All other ASIC must do HQoS in software.

Now as TM try to nest their configuration deeper and deeper, it gets progressively more expensive to perform HQoS and the CPU is pegged.

For DPDK in general, which means all telco grade equipment, each stream will be processed by different core.

In your case by using a IDM, you are basically doing multiple streams, hence on TM side, multiple CPU core is involved and you can saturate your subscribed speed.

How much each core can do HQoS is kind of a trade secret that only vendor and customer knows. Juniper MX Trio 4 ASIC for example, will be pegged at around 5Gbps.

I use Cisco ASR for another example. They quote aggregated bandwidth due to this same reason. Want single stream to max out the port speed? Buy their Silicon One ASIC.
*
You know what, I try all speedtest server, only Digi at Seri Kembangan can saturate my 2Gbps rclxub.gif
user posted image

All other server give crap speed

What do you think?

ibse02.mc give 7ms before
ibse02.rsh now 12ms, that 5ms increase, even stay in Seremban Town, ping make I stay in Kampung, cry.gif

This post has been edited by Anime4000: May 31 2025, 12:30 AM
« Next Oldest · Networks and Broadband · Next Newest »
 

6 Pages < 1 2 3 4 > » Top
Add ReplyOptions
 

Change to:
| Lo-Fi Version
 0.0216sec    0.37    7 queries    GZIP Disabled
Time is now: 26th November 2025 - 02:46 AM
All Rights Reserved © 2002- 2025 Vijandren Ramadass (~unite against racism~)
Removal Request
Powered by Invision Power Board © 2025  IPS, Inc.