Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Terry74]

I have an odd download problem on my FTTC cct.

I've recently moved from BT to Pulse8 as provider, with fixed ip. At first everything was fine, but during the last couple of weeks, I've noticed downloads, probably mostly via Distrowatch.com web site as notification source, have been stopping part-way (~300MB) through the download. These are linux distribution .iso files mostly.(eg Ubuntu-amd64.iso)ie a large file ~2.5GB. Tried also using wget -c "distro.iso" and this too stopped after a similar amount of download, but was able to restart and finish the download in maybe 6 or 8 sections, stopping every 300-400 MB.

It occurs to me that I've only just started downloading .iso files again now the new Ubuntu is almost here, so perhaps it has been this way since the change-over? I just don't know.

My modem/router is usually a VMG8924-B10A in bridge mode followed by an Edgerouter-X router with ubiquiti AP attached although connection to the computer is via ethernet cable from the router.

However, I'm able to download whole BBC programmes off iplayer, using get-iplayer, without any difficulty. These are a similar size or bigger than the distro files and are downloaded at the maximum speed that my cct will attain, approx 30Mb/s.

Now, I have a thinkbroadband quality monitor set up, which pings my router and records the results as a graph. Result included for enlightenment/explanation. The .iso downloads show massive packet loss - 40-50% - the red lines from the top, whilst the BBC downloads show no packet loss at all. The graph to the right of the thick red 35 minute duration line is using a different modem router - another VMG8924-B10A but in router mode just to see if that changed anything. It didn't!! Conclusion, not my equipment.... Unless you know better?

I also have DSL stats running on a Raspberry Pi, so have the graphs from this too. They show no disconnections occuring - maybe something to do with sampling frequency and only 20-30 sec losses - but DLM doesn't seem to notice any disconnections either. I'm sure it would have complained by now if it had.

When these download disconnections occur using either the browser or wget, my edgerouter dashboard shows that the ip has gone missing on the pppoe line of the display. This reappears some 20-30 secs later and from that time, wget will resume albeit manually.

So, questions:-
 
Why am i getting these disconnections?   Where are they occuring?   Are they between the cabinet and me or further back?

Why don't they occur when the source is BBC?

Why the packet loss with wget/browser downloads but not BBC downloads?

Please can someone answer these questions and explain the results I'm seeing?

Terry

Addendum.  identification for BBQM graph - left of red barrier - BBC prog download note no packet loss! - Ubuntu Focal Fossa in many bits lots of packet loss - another go at focal fossa but only first part to prove it still didn't work after disconnecting phones at 1700hrs.
- right of red barrier - new modem/router - download BBC Top Gear at 0940 - 1355 download 512MB thinkbroadband test file stopped at 300MB - small 270MB distro Clonzilla downloaded successfully - download DragonflyBSD iso also slow speed 1.3MB/s

Last 2 don't seem to have caused packet loss?? Is it to do with download rate? Or source? Certainly small files <300MB are no problem

I think I can reproduce all of these results, so not a "one time special"

Just watched "El Camino" on netflix no problem, no packet loss.

Sorry for the long read.

[burakkucat]

I have an odd download problem on my FTTC cct.

<snip>

When these download disconnections occur using either the browser or wget, my edgerouter dashboard shows that the ip has gone missing on the pppoe line of the display. This reappears some 20-30 secs later and from that time, wget will resume albeit manually.

Your local infrastructure reads as "good". I can see no problem with the configuration.

The last two sentences tell of the problem. But why does it happen? 

Why am i getting these disconnections?   Where are they occuring?   Are they between the cabinet and me or further back?

I would say further back into the ISP/CP-land and certainly not something within Openreach's domain.

Why don't they occur when the source is BBC?

Why the packet loss with wget/browser downloads but not BBC downloads?

You are not "downloading from one location" but are using a distributed content delivery network (CDN).

I'm puzzled.    (I wonder if CarlT could offer an option?)

[johnson]

Sounds like your PPPoE session is dropping some time after you saturate the connection.

Have you looked at the ER-Xs logs?

A quick google suggests logs about the pppoe connection should be located at /var/log/vyatta/ppp_pppoe1.log.

If there is anything about dropping the connection after x echo requests I suggest increasing the lcp-echo-interval and threshold. Not sure where to do that with with the ER-X, but in openWRT its under the advanced section for the PPPoE connection. I have to increase these values to avoid the session dropping when saturating the link with my TTB FTTC connection.

[j0hn]

I also have DSL stats running on a Raspberry Pi, so have the graphs from this too. They show no disconnections occuring - maybe something to do with sampling frequency and only 20-30 sec losses - but DLM doesn't seem to notice any disconnections either. I'm sure it would have complained by now if it had.

DslStats detects resyncs seven if they occur between samples.
What you do with the connection cannot effect the DSL like that.

Sounds like 2 possible issues.

The download stopping first.
It could be a faulty line port that keeps resetting anywhere between your ISP and the destination server. There will be many points of failure in-between.

Can you rate limit the download and see if it happens just as often?

The 2nd issue.

When these download disconnections occur using either the browser or wget, my edgerouter dashboard shows that the ip has gone missing on the pppoe line of the display.

The packet loss suggests the connection is being over saturated.
The ThinkBroadband BQM pings come from elsewhere so the fact they aren't getting through backs that up.

I'm not of the opinion the PPPoE session is being disconnected.

The ThinkBroadband BQM shows a continuous connection with packet loss only.

My assumption is the edgerouter dashboard (I've never used it) polls something online frequently to confirm your IP and during the line saturation it is failing to get a response because of the packet loss.

The ISP could check if the VLAN they are using is overloaded. Sometimes common with BTw backhaul.
Might help.

[Terry74]

Thank you for your replies. Just getting round to reading them, had other things to do today.

@burakkucat - these are my feelings, but..... I shall have to look up CDN. Thanks.

@johnson - Unfortunately, I had already taken that system apart and replaced with a 1 box modem/router. I've just looked at the log you suggest with a laptop attached to the ER-X but it only shows 3 entries. For 13th March - created interface...., stopped demon...., started demon...  Again, I need to look up lcd echo interval to see what you are talking about. I intend putting that system back into circuit now, having proved a point, and your reply gives me something to investigate. Thanks.

@ j0hn - As I mentioned, the DragonflyBSD download I did was bandwidth limited and downloaded fully at 1.3MB/s without packet loss (rightermost trace on BBQM) and small files eg Clonezilla.iso (270MB) have no trouble, - one next to it - so looks like you are right about some sort of saturation taking place, but as my FTTC is only 30Mb/s max, I wonder what would happen with FTTP?   When the disconnections occur, my computer (wget) reports loss of internet. I intend pursuing this with Pulse8 on Monday, so thanks for the suggestion about VLAN. I think the backhaul is TT Business but not sure.

Anyway, I will now restore my set-up to as was, so at least 30 mins off line. Don't want to upset DLM too!!

Terry

[ejs]

A BBC iplayer download is probably being downloaded as numerous smaller pieces, which is probably why it's different than a single large file download.

[Terry74]

Just for completeness, updating this old thread.
I didn't pursue this during Covid lockdown so as not to be a nuisance. However, as things got a bit better, I contacted the ISP.
We tried both Dynamic and Static IP's - the dynamic other end in London and the static other end in Manchester TTB centres. No difference.
To cut a long story short, I tried all my collection of routers, in router-only mode, ie ethwan port, Billion 8800NL V1 & V2, Zyxel VMG8924-B10A and edgerouter but the only one that worked was a Zyxel VMG3925-B10B and that was perfect.
@Johnson showed the probable cause of this - the LCP Echo parameters. AND also helped me obtain the supervisor password for the 3925 some time ago.
The VMG3925-B10B has a file "zcfg_config.json" in it's "/temp" directory, presumably copied from the similar named file in the "/data" directory to make it writable. This has parameters X_ZYXEL_LCPEcho:10 and X_ZYXEL_LCPEchoRetry:10
My Edgerouter X has "lcp-echo-interval 5" & "lcp-echo-failure 6" values.
By use of sudo su to be "root" I was able to alter these parameters in "/opt/vyatta/etc/pppoe-provider-template". I set them at 10 & 15 respectively.
Needless to say I didn't discover these files by myself - I simply searched the internet until I found somone who knew what they were doing.
Just recording here in case someone else wants an answer.
Anyway, my Edgerouter seems to now work fine with this TTB cct via Pulse8. No more disconnects.
Regards to all
Terry

[burakkucat]

Thank you for the update. I'm pleased to read that you finally discovered a work-around solution. A good result.

[Terry74]

I shouldn't have opened my big mouth!!
The fault condition is back just as before. Just like someone turned a switch on Monday. I can only download about 350MB before the ppp session stops and I lose the IP.
Only thing that now works is slowing down the download rate using "aria2c --max-download-limit=2500K distro.iso".
That works a treat, or should I say works TODAY!!
However it rather defeats the object of chasing download speed only to reduce it to successfully download something.
Me thinks it's time for a change of ISP.
I wonder if A&A have these sort of problems with their TTB circuits?
Does anyone else use Pulse8 and do .iso downloads?
I'd love to know whether they can download a 1GB test file from thinkbroadband.co.uk.

Regards to all
Terry

[banger]

I was getting this with a TTB Uno line and I replaced the modem (Zyxel 1312-B10A) with a spare and sent back the Asus AC1200G+ and replaced with an Asus RT-AC68U. I was downloading thinkbroadband 1gb file and getting to 800mb and PPP was dropping. I tried several Asus routers to the same effect but it was fairly random and after a PPP session drop would work for ages then do it again. I also noticed that Asus has fixed the WAN stability on one of the recent firmwares of the AC68U. I still dont know if it was the modem or router but it was fixed.

I also experimented with LCP Echo on the Asus and found the lower the number the quicker it would reconnect and resume the download as a workaround and resume audio streaming which runs most of the time.

You have done all the right things and I still dont know the cause, but replacing modem and router fixed it for me.

[Terry74]

Thanks Banger,

I've tried 5 different routers, one being the recommended one as sold by the ISP, and that seemed to work well for a week.

It's good to know that someone else has experienced the same sort of problem as me.

Now what I want is for someone on Pulse8 to download that test file from thinkbroadband, successfully or otherwise.

T