PPP dropped unable to authenticate - 18.00 on 17th Sept 2014
Quote I'm having no issues with my TG582 whatsoever
Neither have I jelv, I binned mine 2 years ago!!
Quote from: Bob Quote from: Townman on 19/09/2014, 15:04
RADIUS servers - is that PlusNet's or BT Wholesale's?
In this latest instant I suspect the latter contributed most significantly.
Whilst the BTw RADIUS would be responsible for giving those customers the BTw IPs, ultimately it would only do so because:-
(a) Plusnets RADIUS were unable to cope with the deluge trying to all at re-connect at once OR
(b) There were insufficient available sessions available on the Plusnet gateways.
I'm getting rather a vague sense of deja vu when it comes to (b)
Whilst you may or may not have sufficient bandwidth, I suspect that much of the problems the other night centers around subscriber session limits.
You were already steering sessions to certain gateways... then something happened just before 6pm and you seemed to drop a pipe.
At this point you would have had circa 30k users competing to reconnect, because you were session steering to certain endpoints, these would quickly have filled, leaving some with nowhere to go.
According to your graphs you have a max of 750k available sessions, yet on average 741k users on-line. That leaves only 9k spare sessions, which isnt enough to cope if one gateway goes down
I'd been sat on a nice quiet shiney new endpoint for the past week (central12.ptw-bng01) and I wasn't affected in the first wave. However I got knocked offline bang on 6:30pm.. implying that ptw-bng01 had very suddenly become overloaded and one of the following occurred:-
(a) The gateway couldn't cope when it reached the max amount of sessions (ie the Juniper)
(b) It was policed
(c) Someone at Plusnet flicked the switch in a manual attempt to re-balance the network.
Ive seen lots of mentions about the first wave of outages, but nothing about what happened later at 6:30. Just what was going on to cause the 2nd .bng gateway to go down too?
Bearing in mind that the bngs appear to have larger hostlinks bandwidth wise and therefore will have more available user sessions... Is it remotely possible that the bng gateways have a hissy when they reach their configured max sessions?
The reason I ask is I remember the JunOS problems years ago with max session during the periods of upgrading from 155Mb to 622Mb centrals and insufficient user sessions. I dont even know if you still use ERX's but I tried to google their session limits these days, interestingly I came across something that seemed to imply that if there were more sessions on a JunOS gateway than directed via the RADIUS server values then the whole lot would drop.
Quote timing out the 172.*.*.* connections sounds like a "smart idea"
That is unlikely to happen there are valid reasons for the 172 IP addresses.. its the fail-over for if the ISPs RADIUS fail, otherwise thousands of routers would endlessly repeat trying to connect.
There are cases when the 172 IPs are needed... ie you need authenticate on the BTw RADIUS to perform additional diagnostic tests
Perhaps a case could be made to reduce the attempts - Im not sure if that is under the control of BT or the ISP now... but I do seem to recall it was BTw - Gawd the joys of coming back to BTw.. I forgot a pile of stuff I learnt years ago and Im now digging deep into my memory banks again.
Quote If would be interesting to know if anyone using a non TG router suffered the same reconnection difficulties?
Yep my TPlink got its knickers in a twist and wasnt able to reconnect. I put my Zyxel back on in the end as it has better monitoring. I strongly doubt that this would be a problem isolated to the TGs
TBH when on LLU I never had a problem with the TG582n.. its rock steady DSL wise and would sit at 3dB for months and months on end no problems with it... and I do mean months - like 9 months plus. Yeah ok the wifi isnt the best.. but its far from the worst. I used a (PN) TG582n with my FTTC modem for many months without issue too.
The thing that most people seem to forget when bitching about its wifi range is that it has a pretty decent DSL chipset which is nice and stable. If I were an ISP Id be wanting to supply a router with a known decent DSLchipset and adsl_phy... you know the important bit of keeping your ADSL connection stable. I keep saying if it really were such a PoS do you honestly think the likes of Zen and AAISP would also be supplying them. They are now a bit long in the tooth and it is perhaps time for Plusnet to start looking at a new unit purely for the wifi side.. but this is going to push costs up.. but I would really hate it if Plusnet chose a router based on wifi capabilities over and above DSL chipset. That would really be fun
Sorry to add to this
... but some one has just brought it to my attention elsewhere, that they thought the available sessions on the bngs may have been reduced. When I went to look.. yep theyve been reduced today. Im pretty certain that at least 2 of the gateways yesterday supposedly supported circa 66k concurrent sessions.
Why on earth would Plusnet reduce the session limits? This doesnt make sense to me unless it really was as I theorised above, that the gateways may fall over and drop all users when they reach a maximum number of concurrent sessions.
Red Herring - ignore this post!
Quote According to your graphs you have a max of 750k available sessions, yet on average 741k users on-line. That leaves only 9k spare sessions, which isnt enough to cope if one gateway goes down
Why on earth would Plusnet reduce the session limits? This doesnt make sense to me unless it really was as I theorised above, that the gateways may fall over and drop all users when they reach a maximum number of concurrent sessions.
Red Herring - ignore this post!
Ah, thanks OldJim.. I guess it is a matter of interpretation and if thats the case - which it probably is reading your explanation - then I apologise for jumping to conclusion that they havent got enough spare sessions.
.....but something still seems up with gateways falling over the other night when they reached a certain figure, or else why would bng2 have also suddenly shed its load at 6:30.. which goes back to point (a) (b) or (c)
... and also if they did have sufficient spare sessions why would so many users be unable to reconnect. The PN RADIUS should be able to cope with a dropped pipe, but obviously couldnt... were they steering too aggressive towards the BNGs
.....but something still seems up with gateways falling over the other night when they reached a certain figure, or else why would bng2 have also suddenly shed its load at 6:30.. which goes back to point (a) (b) or (c)
... and also if they did have sufficient spare sessions why would so many users be unable to reconnect. The PN RADIUS should be able to cope with a dropped pipe, but obviously couldnt... were they steering too aggressive towards the BNGs
Quote from: chrcoluk speedtest on bng2 central10
since the outage every night is congestion.
just hopped back onto a ipv6 gw
central10 ag04
is central 10 on ag04 same pipe as central10 bng02?
sorry to use caps here but after being fobbed off so many times
PLUSNET WHEN WILL YOU FIX THE BNG GATEWAYS TO GIVE END USERS THE CORRECT LINE SPEED ??????????
After this incident I spent ages getting back onto a non bng to recover my 5 meg that I lose when connected to a bng.
I can consistently reproduce the issue that when im on a bng I lose 5 meg throughput. On a non bng I get my full speed.
I have tried talking to plusnet support about this but they deny and fob you off that the BNG gateways don't have an issue. Once spoke to someone in support who was borderline rude about this and, if the situation with the new BNG doesn't improve im off at the end of my contract.
Personally I think they've acquired the bng gateway kit from BT and, well, we all know how rubbish BT broadband/fibre is.