[clug] Something is wrong with my Internet

Tue Nov 8 03:06:02 UTC 2016

Thanks Randall.

Very similar, I've been having dropouts since the weekend, and I (think 
I) am on Reach also.

Sometimes the ADSL sync goes, but other times, the PPP does. I've got 
the modem bridged, so the PPP is on my server.

Glad to know it's not just me, and that bolsters my case when the tech 
comes to solve it.

Would you be willing and able to see if you get the same poor latency 
when you're uploading?  If you need, I can set you up on my server and 
you can scp a file up.  If we have independent evidence of the same 
fault, it reinforces both our cases.

My latest thinking is Buffer Bloat: 
https://www.bufferbloat.net/projects/bloat/wiki/Introduction/

The fix (https://www.bufferbloat.net/projects/codel/wiki/) is in the 
Linux kernel mainstream since 3.6 or 3.8.  Pity both the modem and the 
server are 2.6.X.

Yes, please, I'm interested in your screen scraping code.  I want to 
find the ADSL sync speed programmatically so I can dynamically try some 
traffic control on the server.  I've been delving into the source for 
their web server, but there are some files I can't find (busybox == no 
find command) and so I haven't cracked it yet.  If I can scrape it, 
that'll do.

Tony

On 08-Nov-16 1:11 PM, Randall Crook wrote:
> Hi Tony,
>
> I have been having different problems with my optus ADSL service. May be
> related.
>
> I am in optus' "Reach" service which basically means optus wholesales
> from telstra and I get to pay $20 per month more for the privilege. But
> that's beside the point. I am getting drop outs. Lots of them in what
> appears to be random patterns. Drop outs of up to 2 to 3 minutes.
>
> After a few calls to optus and getting no where fast I decided to to
> gather some statistics. To make it as much on Optus as possible I setup
> their Sagemcom F at st 3864 so they can't complain about who's modem is at
> fault and wrote a script to gather some data.
>
> I also set the modem up to log to a syslog server so I can get an
> understanding of whats happening. I am glad I did as the data seems to
> be telling me the drop outs are not at the ADSL layer like I thought,
> but at the PPP layer. The ADSL is staying solid at 14Mb/s down and 1
> Mb/s up all the time. The PPPoE on the other hand is dropping.
>
> Yesterday I had about 12+ drop outs mostly in the morning all lasting
> about 2 minutes. today so far only a couple.
>
> If your interested I can send you the script I knocked up as it has a
> crude bit of code to scrape data from the sagemcom web interface.
>
> Regards
>
>
> Randall.
>
>
> On 08/11/16 11:04, Tony Lewis wrote:
>> We recently moved house and I've been struggling to get my Optus ADSL
>> to perform like it should.
>>
>> I get sync speeds of 4800 up and somewhere between 160kbps-400kbps
>> down.  I've tried ADSL2+ and ADSL1.
>>
>> It seems the connection is very sensitive to uploads.  For example, I
>> am currently scp'ing a tar file to a hosted server with plenty of
>> bandwidth.  Here's the corresponding ping times to google as things
>> progress:
>>
>> $ ping google.com
>> PING google.com (172.217.25.174) 56(84) bytes of data.
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=1
>> ttl=54 time=52.1 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=2
>> ttl=54 time=43.8 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=3
>> ttl=54 time=46.0 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=4
>> ttl=54 time=44.1 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=5
>> ttl=54 time=44.7 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=6
>> ttl=54 time=45.0 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=7
>> ttl=54 time=52.8 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=8
>> ttl=54 time=52.5 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=9
>> ttl=54 time=48.9 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=10
>> ttl=54 time=674 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=11
>> ttl=54 time=745 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=12
>> ttl=54 time=969 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=13
>> ttl=54 time=1387 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=14
>> ttl=54 time=1736 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=15
>> ttl=54 time=2087 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=16
>> ttl=54 time=2417 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=17
>> ttl=54 time=2909 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=18
>> ttl=54 time=3289 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=19
>> ttl=54 time=3872 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=20
>> ttl=54 time=4558 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=21
>> ttl=54 time=6284 ms
>> 64 bytes from syd09s13-in-f174.1e100.net (172.217.25.174): icmp_seq=22
>> ttl=54 time=6864 ms
>>
>> It gets real horrible real fast.
>>
>> I am pursuing things with Optus, but wanted to see if:
>>
>>   * this was in any way normal?
>>   * there was anything else I could try to fix things.
>>
>> Currently I am using Optus' Sagemcom F at st 3864, which is pretty
>> universally reviled.  I can't seem swap it out because Optus have
>> locked it somehow.  Some think it's through MAC address whitelisting,
>> but logically that implies Ethernet packets, and there ain't none
>> going between me and them.  It's in bridged mode, which gives me an
>> ATM interface on the modem and I can use my linux server as the PPPoE
>> end point.  The ATM interface does have a MAC address, though.
>>
>> I've rooted the modem, so I can telnet in.  It's a cut down Linux
>> (busybox-based) and while I can't figure out how to restart the
>> ATM/ADSL interface, I can reboot it.
>>
>> Blindly in the past I've been scripting a reboot every two hours.
>> This seems to help, but now I don't think so.  I was under the
>> impression that some process on the modem was leaking memory, or
>> chewing CPU, or otherwise being aberrant, because a reboot gave me
>> back good ping times.
>>
>> But now I think that I got good ping times back because any upload was
>> killed or interrupted at least, and so it looked good until that
>> upload resumed, then it goes pear-shaped again.
>>
>> Downloads don't seem to be anywhere near as much of a problem.
>>
>> I've started logging lots of things, including using pmacct to measure
>> my netflows, but I only turned that on last night, so I need more time
>> to get useful data.
>>
>> Any suggestions?
>>
>> Tony
>>
>>
>