Print

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

[Weaver]

In the early hours of Saturday the 15th there was a clap of thunder which woke us up. Janet scrambled to pull modems out, but it was too late. One ethernet cable was blackened, and the VLAN switch that muxes the four modems together was wiped out. Had to wait until Monday to talk to AA. AA sent us a new switch, no joy, then another new switch and a new Firebrick FB2900 which arrived yesterday, Friday 28th. Still no joy. All ethernet cables and RJ11 cables were replaced with brand new ones straight out of their packets but no luck.

All week we had been running on 3G USB ‘dongle’ failover and 4G AA iPad and EE iPhone SIMs, racked up a bill of £112 which AA refunded. Tonight I got an email from AA - the engineer who had been working with us all week had found the answer; AA had sent us two MUX switches with bad config in them. I don’t quite understand how the small MUX switch config came to be wrong, because there is a config file on the AA support website and I would have though that they could just use that?

So last night the mystery was resolved. Bad news about a week of semi-downtime, or at any rate restricted use, ie. having to be very careful with 4G/3G. I had accidentally downloaded a movie before I realised that we were running in failover 3G mode, racking up a huge bill. AA are very good and send out a warning email when your expenditure goes over a certain limit though.

All week I just couldn’t work out what was going on. I told myself that AA couldn’t possibly have sent me two bad MUX switches, but that was indeed what it was. I tried my FB2500 too, with appropriate configuration tweaks but no joy with that either so it seemed nothing worked. The modems were known to be good because BT remote modem status tests executed by clueless.aa.net.uk showed the modems were up, but I couldn’t ping the modems (this was because of the bad MUX switch's config). So it was driving me mad - known good fb2500 (bar any possible fb2500 config mistakes of mine), good modems, all new cables throughout and two new AA mux switches.

Somehow our AA engineer remotely fixed the mux switch config - must have been using NAT in the Firebrick as the switch has an RFC1918 admin IP address. I had created a remote admin login account for AA’s use so they could administer the Firebrick remotely (and I have set up a firewall hole for AA staff to allow them to access anything in my world).

Janet wants to send back the new FB2900 firebrick which arrived today - we don’t have any money for such things.

At least the mystery is finally solved. But what a nightmare. How many lightning strikes with damage does that make it so far this year?

Later on on Friday evening, I managed to unfix the fix in the mux switch’s config. Speculation: perhaps Janet power-cycled the unit, and the amended config had not been ‘saved’, that is, not committed to non-volatile storage ? In support of this speculation, Burakkucat tells me this behaviour is not unknown - where changes made to a unit’s config are not committed until some special ‘save’ command is issued - a design that helps prevent locking yourself out easily due to config-editing mistakes, because you can easily recover by simply rebooting the unit.

I discovered one of the modems is bad so I replaced it. That’s a small job for me on Saturday, configuring a new ZyXEL VMG1312-B10A.

[Weaver]

Many thanks to Burakkucat who kept me sane all week and acted as a sounding board for me at all times. Some times just trying to explain things to someone else helps clarify your own understanding.

[burakkucat]

<snip>
At least the mystery is finally solved. But what a nightmare. How many lightning strikes with damage does that make it so far this year?
<snip>

Too many. I've lost count . . . 

Many thanks to Burakkucat who kept me sane all week and acted as a sounding board for me at all times.

You are welcome. 

(There's no way that I could just "drop in", to assist with a fix.)

[burakkucat]

With reference to the query in the subject line, I've spent a little time in adding the month & year to each topic found.

On the first page of the reverse chronological order of subjects, to which I have attended, four were found just going back to last October (2019). 

[Weaver]

Definitely never had a year as bad as this.

[Weaver]

Burakkucat and I were discussing the possibility of using a VLAN-aware switch that has an SFP slot in it so that I could use an optical link to the FB2900’s SFP port.

There are problems with this: I wouldn’t know how to configure the VLAN switch for muxing; the cost of the VLAN switch given that it is a sacrificial lamb/goat/lobster and has to be replaced often due to lightning zap; I don’t know anything about SFP modules or optical links either. But that’s all.

I did see a TP-LINK SFP-to-gigabit-copper converter. That solves the cost problem partly, in that I can use a cheap MUX switch, and it solves the ignorance problem in that I can continue to use the current familiar ‘supported’ ones. It adds the cost of the converter that’s all. So not a bad plan.


I suspect Burrakucat and CarlT - hi Carl - could help me out here though ?

[Weaver]

Somehow I need to fix the other MUX switch. Well, the first MUX switch needs re-fixing first. I think the first MUX switch’s admin i/f was given an IP address by our AA engineer by tweaking the Firebrick’s config so as to use NAT to make that switch accessible.

[burakkucat]

I have been thinking about your idea of optical isolation. The equipment either side of the optical link will require power.

(Q) From where will that power be obtained?
(A) From the mains supply in the "Weaving Shed".

Therefore both sides of an optical link will be electrically connected via the mains supply. As far as I am aware, the average switch-mode power supply is doubly insulated. A significant potential difference would be required before there is breakdown between the low voltage DC output and the AC mains input to the device. So how much protection would an optical link provide? Is it possible that an induced high voltage spike on one line, destroying a modem and its power supply, will be sufficiently dissipated via a multitude of paths through the mains wiring not enter the Firebrick through its power supply?    I have no idea.

[Alex Atkin UK]

With reference to the query in the subject line, I've spent a little time in adding the month & year to each topic found.

On the first page of the reverse chronological order of subjects, to which I have attended, four were found just going back to last October (2019). 

That's one thing I hate about this forum, the search facility seems woefully basic.  At least being able to sort by date would be insanely useful.

[Alex Atkin UK]

I have been thinking about your idea of optical isolation. The equipment either side of the optical link will require power.

(Q) From where will that power be obtained?
(A) From the mains supply in the "Weaving Shed".

Therefore both sides of an optical link will be electrically connected via the mains supply. As far as I am aware, the average switch-mode power supply is doubly insulated. A significant potential difference would be required before there is breakdown between the low voltage DC output and the AC mains input to the device. So how much protection would an optical link provide? Is it possible that an induced high voltage spike on one line, destroying a modem and its power supply, will be sufficiently dissipated via a multitude of paths through the mains wiring not enter the Firebrick through its power supply?    I have no idea.

From the various YouTube videos I've seen of teardowns, most switching PSUs have pretty poor isolation as the transformer only has the insulation of the wire to separate the primary and the secondary windings.  That said, its a better insulated path than an ethernet cable.  If the surge made it to the mains, everything in the Weaving shed would be fried by now rather than one or two modems.

The issue seems to be that it takes very little current to fry things directly hanging off the phone lines, so the optical isolation I believe would still help.  Once it reaches the mains, it should be shunted to ground if there are surge protectors in the house, reducing the impact from there on.

I can't think of any foolproof solution, obviously the idea is to keep the cost impact down as much as possible.

[sevenlayermuddle]

I can see an argument that, if the conversion from copper to fibre consisted of an old fashioned light bulb shining into glass, it would be relatively immune it common mode impulses.  Such impulses, being common mode, wouldn’t be transferred to the glass, and neither would they overload the filament.

Of course, conversion is not done by old fashioned filament bulbs.  It is done by, typically, LED devices after active amplification and processing of the signal received from copper.

My gut feeling would be that, without grounded over-voltage protection, the circuits that drive the LEDs would quickly die in the event of lightning surges.   The modem at the other end  of the glass would obviously survive, but the data connection would still be permanently lost pending replacement of the damaged hardware.

[Weaver]

Can anyone tell me what the resistance per unit length of cat6, cat7, cat7a or cat8 network cables is? I’m using 1m or 2m cat8 cables to interconnect the modems to MUX-switch and MUX-switch to firebrick etc. No good reason at all, but I don’t want to have a lot of the older standard cables lying around polluting the pool of cables if I ever manage to get a 10Gig main switch.

[Ronski]

That's one thing I hate about this forum, the search facility seems woefully basic.  At least being able to sort by date would be insanely useful.

You can, just change the search order to oldest topics first or most recent topics first.

[Ronski]

What about purchasing some insurance backed surge protectors? Even if they don't work you should be able to claim on their insurance, although I've no idea how good they are at paying out.

[Alex Atkin UK]

You can, just change the search order to oldest topics first or most recent topics first.

Dang it, its that old stupid thing where quick search doesn't let you change the advanced options AFTER performing the search.  Still dumb, but at least there is a way.