[apologies for the length of this post; seems to have grown into a monster
]
I have a 3G USB ‘dongle’ interface and I wished to use this with my Firebrick router, which has a USB port on it intended specifically for this purpose. I wanted to position the dongle in the window where it would get a direct view of the basestation across the glen to the east/northeast. So I bought a 5m long USB cable which is - I think, can’t remember - described as ‘active’. I had the idea that 3m was the maximum length but I think that that was still in old money and it’s maybe 5m for USB 2.0, so that’s why I bought the active cable.
I would have thought that such a device would need an external DC PSU input, no? This has none, so I’m assuming that it’s just simply designed to live off the ample power available from the USB cable. Is that usual?
The dongle connection has been monitored by AA’s usual PPP LCP ping servers (FB60xx series) and in the past the connection has repeatedly been seen as down. I don’t know whether this is sometimes a problem with 3G or the mobile network’s connection back to AA, but there has certainly been a problem with the 3G dongle as when the link to the dongle is seen by AA as down, I have gone to try Firebrick FB2900 and asked it what’s up, and the FB2900 says that the dongle has gone out to lunch. The USB port is shown as up, but
iirc no dongle device seen on it, and certainly no 3G PPP sessions shown on the USB i/f.
So badness with either the dongle, the FB2900’s USB i/f, the dongle driver, or the USB cable. So the easiest way to eliminate one of those is to get rid of the active cable and plug the dongle straight into the FB2900, which is the usual way it’s done, and is the way that Firebrick Ltd tested the system. This is a known good model of dongle, recommended by AA and used by them in their testing so there shouldn’t be any bugs in the dongle itself or the FB2900’s code because this is then all a known setup. The dongle should still have a line of sight to the basestation given where the FB2900 is; on the south wall of the upstairs office, facing north, bay window on the east side of the room, so there’s a line of sight to the northeast just right for the basestation and dongle. I’m wondering if the FB2900 USB port doesn’t supply enough power for ‘extras’, ie anything above one dongle, so perhaps not enough for a greedy active USB i/f regenerator device at the far end.
Assuming the active cable has a regenerator at the dongle end, won’t it need a second one at the FB end too? I haven’t studied it.
Next bad thing. I bought a small external antenna from somewhere, I forget - either eBay or Amazon. This plugs into a tiny port on the dongle. I’m wondering if this was absolutely rubbish and the dongle would be better without it. I asked Janet to unplug the antenna, can’t remember whether she got around to it or not - will have to ask her. I also iirc talked to her about getting rid of the active cable, but again, am not sure whether she has done so yet or not.
Anyway, the 3G link has been problem free for a long time, which is a big change seeing as before it would be down/up/down several times in the day, down for a couple of hours approx each time. And this improvement may be because Janet has made some of these physical changes, will have to ask her. She’s very unwell at the moment, recovering from surgery, so I haven’t pestered her with unimportant things such as asking her to inspect dongle connections. Several years back, the 3G link would be fine for months on end and then go down, for reasons unknown. I had a moan about it with AA, and wanted them to set up a long term test bed, but I did warn them that I had to wait several months for mine to fail, and for all I know it could be something to do with my basestation which will possibly be different with theirs. Anyway they weren’t immediately willing and I didn’t push. I can’t remember, but I suspect that back then such a link down state would also be associated with a no dongle and/or no 3G-session being visible at the FB2900. It could be that now we’re back to that improved but still unacceptable level of reliability.
I say ‘unacceptable’ because it shouldn’t be happening, as this is a critical link used for failover and must always be there when called upon, and also because there’s currently nothing to tell me that it has failed. AA could do this but they have (somehow) set dongle and other mobile connection objects in clueless to have no notification events associated with them. This is fair enough for 3G/4G/5G interfaces used in portable devices such as an iPad as there would otherwise be lots of bogus irrelevant notifications all the time whenever the host goes out of signal, goes into deep-sleep or ‘off’ or off mode. There need to be explicit flags visible to control this behaviour: suggest a fixed-position/movable flag and/or a notifications y/n flag. The former, fixed-position, state would cause notifications to be delivered and the device would always be expected to be in signal and up or else it’s considered that something bad has happened. I might suggest this to AA as it’s an easy change.
If I do a bit of research or just ask AA, I think that there might be a way to get the FB to generate notification events when a dongle goes down. It can possibly turn notification events into emails, I think.
I wanted to do IP ICMP ping monitoring using the servers at
https://f8lure.mouselike.org/, but a traceroute into the IPv4 address associated with the FB2900-end of the 3G link seems to show a route that is insane, so ping monitoring isn’t currently possible. The route comes into my firebrick at the normal WAN i/f; this will be DSL now, as normal, in the non-failover state; from there the route goes to the designated IPv4 for the dongle at the downstream end of the 3G link at the firebrick; next it goes back up the 3G link to an AA router and then reverses heading back downstream to the dongle again, so a routing loop! My fault definitely. AA will reroute traffic when all my DSL links fail and direct traffic into the ‘secondary routing’-marked links in clueless.aa.net.uk, which in this case are set to be the 3G link. The route downstream to the
explicit FB-end of the 3G link should as I see it always be down the 3G link. This is what I need to be able to ICMP-ping the 3G link to test it. I’m pretty sure that I have messed up the tickboxes for primary and ‘secondary routing’ controls for each link in clueless and am asking AA to have a look and tell me what a wally I’ve been.
If I go to a 4G dongle - AA sent me one free, but Janet may have ‘filed’ it, so that could be trouble- then this might well be the type that is an ethernet device, and not a PPP-speaking modem, but a stupid NATing router. The enforced NAT is a disaster, so I decided not to use it and didn’t have the mental energy to investigate the possibility of reconfiguring it to be a straight modem (and didn’t have the required tools anyway). But now I have read that L2TP services are free to all AA users. I thought you had to pay for L2TP, but that seems to be just for people who are not current customers. I could use L2TP to route all known traffic to my existing IPv4 /26 and my IPv6 /64 (out of a /48) so I would get seamless failover provided I get the reduced MTU right and make it a permanent restriction. I could let IPv4 fragmentation and keep the MTU at 1500 and maybe the upstream end of the L2TP might reassemble all fragments, maybe not, but AA do growl at you if you pass fragments through.
So a lot of varied questions - quite a few things that I need to ask AA about as I don’t know what I’m doing, but perhaps you could put me straight in some of them.