Troubleshooting ADSL problems - not the usual (I think)

[RoganDawes]

Hi folks,

My problems all started after a direct lightning strike took out my surge protector. Since then, it has been up and down at random (frequent) intervals, and I'm trying to understand what is going on. My set up is as follows:

DLink 2500U ADSL modem in bridged mode, speaking to a TP-Link WR1043ND running OpenWRT.

Note: I have replaced the ADSL modem with an identical spare, so I don't think this is related to my equipment. Also, the surge protector has been replaced too.

The OpenWRT router is logging everything to an internal server, so I have some historical logs to work with. The 2500U is also logging to the same internal server, so I get stats from that too.

I am also monitoring the line stats of my 2500U, so can see things like the bit errors, etc, attenuation, and so forth, and especially, things like the ADSL sync dropping and renegotiating. This is not happening, that I can see. To manually verify the logs my modem generates when the actual ADSL connection goes down, I yanked the phone line, and got:

Feb 26 09:52:41 adsl kernel: ADSL G.994 training
Feb 26 09:52:48 adsl kernel: ADSL G.992 started
Feb 26 09:52:53 adsl kernel: ADSL G.992 channel analysis
Feb 26 09:52:59 adsl kernel: ADSL G.992 message exchange
Feb 26 09:53:00 adsl kernel: ADSL link down
Feb 26 09:53:05 adsl kernel: ADSL G.994 training
Feb 26 09:53:12 adsl kernel: ADSL G.992 started
Feb 26 09:53:16 adsl kernel: ADSL G.992 channel analysis
Feb 26 09:53:23 adsl kernel: ADSL link up, interleaved, us=507, ds=4096
Feb 26 09:53:23 adsl kernel: ADSL2/ADSL2+ connection

These are the only entries indicating an actual ADSL line renegotiation, otherwise my line is pretty solid at 4Mbps actual sync rate.

However, I cannot get a PPPoE connection to remain stable. Sometimes it will stay up for "0.2 minutes", and sometimes it will remain up for 12 hours (few and far between, though).

Obviously, I reported it to Telkom, but it's a bit difficult for them to see the problem, given that it is intermittent. So they came around while I was out, "checked the line", and reckon it is fine.

Note: These syslog files cover log entries for the past week.

Disconnects by day:

# zgrep -h pppd syslog* | sort | grep -v "Sep  " | grep "Serial link appears to be disconnected" | cut -f1-2 -d " " | uniq -c
      4 Feb 19
     23 Feb 20
     35 Feb 21
     50 Feb 23
     29 Feb 24
     51 Feb 25
    148 Feb 26

As you can see, it is not getting any better :-( Also, this does not reflect any manual disconnection or bringing down of interfaces during debugging, this message is emitted when pppd detects a link failure.

I thought of drawing stats regarding what time the disconnects happen:

# zgrep -h pppd syslog* | sort | grep -v "Sep  " | grep "Serial link appears to be disconnected" | cut -c8-9 | sort | uniq -c
Count per hour
      1 00
      8 01
     45 02
     51 03
    101 04
     53 05
     23 06
     21 07
      7 08
     10 09
      1 10
      2 11
      6 17
      6 18
      3 21
      2 22

So it kind of looks like it is reasonably stable during the day (no disconnects between 11:00 and 17:00), and totally crappy during the early morning.

In order to confirm whether this is my ISP or a general problem, I established a non-routed connection using the Telkom Guest account:

# zgrep -h telkomguest syslog* | sort | grep -v "Sep  " | grep "lost the connection" | cut -c8-9 | sort | uniq -c
     10 02
     10 03
     24 04
     15 05
      2 06
      3 07
      1 08
      3 09

Note that I only configured this last night, but it suggests that this is not ISP-related either.

Any suggestions?

 

[giggity]

Connect your PC directly to the modem and see if the same problem persists. Also, get the ADSL line stats from the router and post here.

 

[RoganDawes]

Hi, I tried connecting my PC directly to the modem, and got the same results (crappy connection) as I did going through the router. What's kind of odd is that the ADSL modem does not report PPPoE disconnects nearly as frequently as the router does. That may just be because it doesn't give up after 5 missing heartbeat packets, though.

Here are the modem stats:

Statistics -- ADSL

Mode:	ADSL2+ 
Line Coding:	Trellis On 
Status:	No Defect 
Link Power State:	L0 
 
 	Downstream	Upstream
SNR Margin (dB):	11.5 	16.5 
Attenuation (dB):	38.0 	18.4 
Output Power (dBm):	-1.7 	0.0 
Attainable Rate (Kbps):	11632 	1172 
Rate (Kbps):	4096 	507 
MSGc (number of bytes in overhead channel message):	59 	12 
B (number of bytes in Mux Data Frame):	128 	15 
M (number of Mux Data Frames in FEC Data Frame):	1 	1 
T (Mux Data Frames over sync bytes):	1 	4 
R (number of check bytes in FEC Data Frame):	0 	0 
S (ratio of FEC over PMD Data Frame length):	1.0000 	0.9922 
L (number of bits in PMD Data Frame):	1032 	129 
D (interleaver depth):	1 	1 
Delay (msec):	0 	0 
 
Super Frames:	184984 	176767 
Super Frame Errors:	122 	0 
RS Words:	0 	0 
RS Correctable Errors:	0 	0 
RS Uncorrectable Errors:	0 	N/A 
 
HEC Errors:	30 	0 
OCD Errors:	0 	0 
LCD Errors:	0 	0 
Total Cells:	29042878 	594365 
Data Cells:	6874832 	16930 
Bit Errors:	0 	0 
 
Total ES:	181821 	0 
Total SES:	176454 	0 
Total UAS:	194 	0

 

[giggity]

Sounds like your hardware is faulty (not the router/modem). I used to have the same problem after a lightning strike. The little grey boxes where the line splits or converts to a phone jack were completely singed. Surge protection doesn't protect these parts -- only what's after it. Have a look inside the boxes and see how things look. You might see clear damage. Contact a Telkom technician and tell them you had a lightning strike and there is damage which needs to be repaired/replaced. They replace lightning-damaged parts for free.

 

[RoganDawes]

Ah, I'll do that, thanks. Hadn't thought of checking the sockets. They are ones that I installed myself, so I don't think Telkom will be interested ;-)