automatic plusnet gateway swapping

is it normal for plusnet to kick users to swap their gateways?

noticed this morning all my ssh sessions dropped and I see my ppp was dropped at 6am to put me on a different gateway.

Jan 1 05:45:08 TomatoASUS daemon.info pppd[8492]: LCP terminated by peer
Jan 1 05:45:08 TomatoASUS daemon.info pppd[8492]: Connect time 746.9 minutes.
Jan 1 05:45:08 TomatoASUS daemon.info pppd[8492]: Sent 469091194 bytes, received 2974499477 bytes.
Jan 1 05:45:12 TomatoASUS daemon.notice pppd[8492]: Connection terminated.
Jan 1 05:45:12 TomatoASUS user.info redial[8493]: WAN down. Reconnecting...

Edited by Chrysalis (Wed 01-Jan-14 12:10:06)

Happened quite a few times on adsl2+ yet to see it on fttc

It could have been BT (in fact that is more likely).

I hope it isnt a regular occurance given broadband is supposed to be 'always on'. Been dropped will drive me nuts as I have established connections that should last for days/weeks at a time.

There's nothing about it in the Plusnet Service status emails. Have you set your account to be sent those? They are very useful.

They haven't enabled IPv6 for you have they?

no that would need to me to use new login details I assume.

Lets see what happens from now on first, it may be related to me been a new customer as I have noticed my rdns has changed from yesterday and also my throughput is 2mbit down suggesting a profile change. Before I maxed at 7.7mB now its 7.5.

so basically none of you guys had a ppp drop last night and you usually have long uptimes?

Edited by Chrysalis (Wed 01-Jan-14 16:34:37)

I get this occasionally, maybe every couple of weeks, usually early in the morning.

In reply to a post by kasg:

I get this occasionally, maybe every couple of weeks, usually early in the morning.

This.

In reply to a post by Ignitionnet:

This.

What?

In reply to a post by Chrysalis:

I hope it isnt a regular occurance given broadband is supposed to be 'always on'. Been dropped will drive me nuts as I have established connections that should last for days/weeks at a time.

In my experience it is. Plusnet occasionally have to balance customers between endpoints or there's a hit on the BTWholesale network.

It's rare enough and the outage brief enough that I haven't noticed until after the event, however I am not running any kind of service monitoring from home and am not doing anything that can't be easily reconnected.

For this kind of thing the safest bet is to avoid BTWholesale entirely or use an ISP that runs with more headroom on their endpoints and/or is less responsive to imbalance.

That was a way of saying 'me too', Kevin.

In reply to a post by Ignitionnet:

That was a way of saying 'me too', Kevin.

Ah sorry, didn't follow the lingo.

ok I can deal with it if its every couple of weeks or more.

indeed it shouldnt be cutting out all sessions if its a short outage, for instance I have in the past been able to do a full modem reboot whilst preserving connections.

so part of the problem is plusnet cutting out my "always on" connection but another part seems to be my router isnt gracefully handling the ppp been disconected

this is further down in the log

Jan 1 05:45:12 TomatoASUS daemon.info pppd[8492]: Terminating on signal 15
Jan 1 05:45:12 TomatoASUS daemon.info pppd[8492]: Exit.

so a sigterm is been sent to pppd itself which I think will terminate all live connections.

tomato firmware has some nice features but it has various things implemented poorly on the underlying network handling which I think have been left alone for years since the original fork of tomato firmware stopped development.

just 2 seconds later the new ppp session was up

Jan 1 05:45:13 TomatoASUS daemon.notice pppd[27915]: Connect: ppp0 <--> vlan2
Jan 1 05:45:14 TomatoASUS daemon.info pppd[27915]: CHAP authentication succeeded
Jan 1 05:45:14 TomatoASUS daemon.notice pppd[27915]: CHAP authentication succeeded

so a total of 6 seconds outage was quite devastating which I think was caused by the pppd been terminated. I have had 30 second outages far less noticeable.

yeh I get signal 15 disconnections on my old adsl2+ early hours of the morning

I think we both have similar lines judging from the sync speeds you've published.

I've had very few disconnections that I haven't initiated. My longest up-time is about 3 weeks but I just have to tinker with something... I've been forcibly disconnected from Plusnets end about 3 times in the last 12 months. Including the time when they re-created my account.

One thing I've noticed personally which you may want to test for yourself is the gateway you connect to. I get much better throughput and latency on Plusnet gateways that start PTW or PTN... Some of the PCL gateways drive me nuts... I stick with PTN and PTW and I'm usually ok.

In reply to a post by Chrysalis:

is it normal for plusnet to kick users to swap their gateways?

noticed this morning all my ssh sessions dropped and I see my ppp was dropped at 6am to put me on a different gateway.

If there's a change to your BT IP profile following a resync event then the profiling at our side is normally adjusted to match. When this happens it typically forces a PPP disconnection/reconnection. This might have been what happened.

Other possibilities are planned/unplanned maintenance at our side, across the BT Wholesale Network or an unexpected outage of some description.

Regards,

Hi

there was no resync event, not even during the migration.

In reply to a post by Chrysalis:

so a sigterm is been sent to pppd itself which I think will terminate all live connections.

Depends - whether sigterm or just while renegotiating after an lcp term ppp0 is going to disappear.

If you are doing nat using MASQUERADE then iptables will clear the conntrack state in the expectation that a new ip will be obtained.

If you are doing NAT and you have a static IP you can work around this (router firmwate permitting) by using SNAT instead of MASQ.

Another issue separate from NAT - when ppp0 disappears your box will send back an ICMP net unreachable to the LAN. You can work around this by blocking them with iptables in output.

Of course if ISP sends them back to sender while you are not connected that's out of your control - but when I was with Enta I had a set up with SNAT + blocking locally generated lan bound ICMP net unreachables and I could kill pppd while downloading and it would resume once I was back up.

thanks for the info its interesting.

Looking at the live iptables rules on the router it is using both SNAT and masquerade.
However SNAT is only used on the lan ips masquerade is used on 0.0.0.0 meaning what you said is probably happening. Whether or not I can manually change this with it sticking (at least between reboots) is another matter tho.

For ppp persist is configured so the sigterm defenitly shouldnt be happening.

I think you may well be right regarding iptables but given there is some rules set with my wan ip I believe these rules are probably getting updated everytime ppp connects and if thats the case me manually changing them probably wont be effective, but I will try to replace the postrouting masq with snat and see what happens when I cycle ppp.

--edit--

Ok I found a setting in tomato for SNAT, enabled it so dont need custom rules. Manually disconnecting ppp still terminates all connections so I did a test with a low LCP timeout, pulled the wan cable for 30 seconds, put it back in, log shows ppp was lost and started again, and my ssh session held, so hopefully this does the trick.

Edited by Chrysalis (Sat 04-Jan-14 01:58:26)

Well tomatoasus I had trouble getting ipv6 working correctly, although I got it working the things I had to do I thought was very silly so am preparing a report to both the 2 guys maintaining forks (although I think wont get fixed), and I changed back to the merlinasus firmware which is much more compliant with RFC's. The problem is this firmware is masquerade with no option to use snat, so will be contacting this dev to get it implemented though he is much more active than the tomato guys.

this is only an issue on ipv4 tho, from what I can see on the ipv6 iptables the problem will be ipv4 side only.

It's possible that just disabling nat by gui and then making manual rules would let you do snat.

Of course I don't know your firmware at all. When I first did this on a modem/router it was using routertech firmware which didn't have snat in GUI, but did have it in kernel. Later I got a /29 and and started doing NAT on a separate PC I used as a gateway.

WRT still loosing connections when pppd is killed, maybe something is happening in the scripts which by default are /etc/ppp/ip-up /etc/ppp/ip-down (routertech didn't use those on the version I used back then).

yeah although I expect that wont be too easy either, all these custom firmwares, asuswrt etc. one thing that cant be customised is the iptables rules and the scripts they use seem hidden away. They write temporary config files instead of permanent one's.

I will leave it alone for now.

Post deleted by AndyHCZ