Week of 20th August, 2007

Last Week

Next Week

Monday, 20th August

A couple of weeks ago, I commented about uptimes on the Vigor router, and my suspicion that the link uptime was displayed modulo 1024 hours. I can now state that this is not the case - the displayed uptime counted through 1024 hours with nary a hiccup. On this basis, it's unlikely that there will be any change until the appropriate snmp timeticks variable rolls over - and that's a 32 bit counter, incremented 100 times a second, for a period of just over 497 days.

That said, the wireless access point - a Linksys WRT54GS running DD-WRT V23 SP2 - is giving slow, bursty throughput, at an uptime of about 186 days. Maybe I should restart it.

And despite the issues with total data throughput (and my lack of knowledge of the size thereof) night-time downloads can occasionally peak to 4Mbit/sec or better. My speed problems thus seem to be mostly contention based. But this is to be expected - domestic ADSL provision here in UK is on the basis of a contention ratio of 50:1. If you want better, you spring for a Business-class connection, and you pay for it.

Talking of paying, I can get the new, reduced monthly rental of UKP22.95, down from UKP24.95, if I sign a new contract, which would lock me in for 12 months - at present, my contract is old enough that I can escape from it with 30 days notice. Not that I'd want to escape, but why the new contract, just to get the new, reduced, price?

And I've ordered the various electronic bits from CPC. Should be here by Wednesday.

I'm noticing increased, lower-level, traffic on the ADSL link. As far as I can tell, it has to be one computer that's doing it, and that's Katy's former desktop, that is now installed in the kitchen, for Jane's use. And the traffic to that machine is driving Tux's ntp server nuts. I think I'll hve to instrument Jane's machine for snmp.

Later: No need to do anything to Jane's computer - it's Jenny. Running Bittorrent - the Windows BitComet version. Should have realised from the behaviour of Tux's ntp graph - bittorrent plays hob with ntp, especially when maxed out.

Back to Daynotes


Tuesday, 21st August

Rats! The router restarted about 21:00 tonight - that's kippered the uptime, which was in excess of 4400 hours. And, of course, the link dropped, as well, and didn';t come back up until I tried to go online after work.

My first thought was a power outage, but the router shares an APC UPS with Tux, and he didn't restart.

Still, I shouldn't complain - 200+ days is good. I can live with that. If it hasn't clobbered my comics downloads...

<pause>

which it hasn't.

Overtime today, and the mixture as before. Lots of hurry-up-and-wait. But whatever, extra green stamps are not to be refused.

And CPC delivered today, as the 'phone agent said. The order complete, but one item is wrong - either a transcription error when I read the numbers, or a picking error. I asked for a USB battery box that takes 4no. AA batteries, but I was sent the similar device that accepts AAA batteries.

The idea here is to have a device that can provide USB power to recharge something like an iPod - which Jane has just bought for herself, in addition to the ones our daughters have. And it has resulted in considerable fulmination from Jane, who is computer-illiterate, and doesn't understand the conventions - and she is vocal in her complaints. Trouble is, Wind0ws desn't do things the way she thinks they should be done - I will admit that some of the things she complains about are stupid, but her logic keeps leading her astray. She doesn't think like a Wind0ws user. And inanimate things misbehave for her anyway.

23:30. That's worrying - the router is spontaneously rebooting. And it takes a while to reconnect, which means some things time out - like mrtg. Looking at the ADSL attenuation plot in mrtg, which is the best indicator of link drops, shows that problems started at just after 18:00.

Midnight: I suspect that the DSLAM has fallen over - there doesn't seem to be any bitstream. No biggie - there's nothing irreplaceable until tomorrow evening. Mail will stay on Demon's smarthosts, newsgroups will be held on the news server until I can connect again, so it's just the comics download.

Back to Daynotes


Wednesday, 22nd August

Well, I now don't think it's the DSLAM - the problem continued overnight. I very much doubt that even BT could be so cack-handed as to disregard the multiple alarms that would be ringing at my local exchange and it's parent and peers. Besides, simple loss of ADSL bitstream shouldn't cause the router to reboot.

No, I have a dud router. It's 14 months old, so it's out of warranty - I need to buy another. Unless I reinstall the Netgear DG814 and give up snmp monitoring of datarates. Nah. New router it is. I'll get another Vigor 2800 - this one has worked well for 14 months.

So I checked prices at work - and Transparent Communications are as cheap as any, they have stock, and will dispatch tomorrow. Previous experience with them suggests delivery on Friday. Good enough.

Other than that, there's nothing to say about work - it was the (quiet) mixture as before. But the offer of overtime shifts for next month has been circulated, and I've got my bid in - successfully. Trouble is, 3 out of the 4 shifts I asked for will be nights - nobody is taking leave on 10 or 12 shifts. Just like me...

At home, the comics download had failed, but running the script manually (during a time when the router was working) fixed the problem.

This is a plot of the loop attenuation, obtained by snmp from the router. You can plainly see the many link drops - each of which is due to a router restart.

This shows Tuesday and Wednesday, with time increasing to the right (the mrtg 'growright' option)

Back to Daynotes


Thursday, 23rd August

Transparent are to be congratulated - this morning came a knock on the door. It was the router, courier-deivered. Kudos to all. But I can't install it until this evening - no time before work.

And work was as before - quiet.

In the evening, I unboxed the router, and proceeded to copy all the settings from the old to the new. And after a few mis-transcriptions, it all works. Router in service at 23:00 or so, full snmp reporting enabled at 23:20 - bear that in mind when you read the graphs.

And the latest firmware (V2.7.1) for these routers supports IP Binding - Draytek's name for static IP address allocation via DHCP, which I will use.

The new IP address map here will eventually read -

192.168.x.1 192.168.x.63 servers and infrastructure
192.168.x.64 192.168.x.127 workstations
192.168.x.128 192.168.x.191 visiting machines
192.168.x.192 192.168.x254 unused

where x, the third octet, is confidential. All will share the same netmask - 255.255.255.0.

Initially, everything will go into the visiting section, and I'll move them manually as and when I see them. If I can do it, I may well implement blocking of IM applications for IP addresses in the visiting section,and leave wireless laptops in that group - that's if I get cheesed off with certain people sitting in front of the TV, channel-surfing, IMing and texting all at the same time. BOFH here I come...

And it crashed at 23:50... But this is the original power brick - that's the next thing to change. Tomorrow, if necessary.

Back to Daynotes


Friday, 24th August

After several more crashes, I finally got fed up, and during a router restart at about 16:30 today, I swapped out the mains brick - the old one, that I didn't change yesterday - for the new one. And the router started up and locked faster. Hmmm... now to see what happens.

19:00 Another reboot, or at least a link drop. Let's check the router uptime... 40 minutes (at 19:45) It was a reboot.

I don;t know what is causing this. I've changed everything that could be responsible - except the UPS. Or is there something in the ADSL bitstream that's causing a router crash? Or... is the microfilter playing up? My filter is at the telco demarc, and there's a separate cable from there to the router. Or, for that matter, is it the cable?

Ghod, this gets worse. Anyway, I can put in another filter - but Draytek don't include one in the box. In any case, why should anything on the line side of the router trigger a reboot? All I can do is systematic replacement of elements, seriatim. Hopefully, one of those will cure the problem.

Looking at the router's downlink rate statistics, I can get a sense of the average throughput. mrtg is claiming that my 2-hour average downlink rate, over the last month, has been about 120kbit/sec, which converts to about 1.3GB/day, comfortably under Demon's 60GB/30 day cap, which corresponds to 2GB/day - although Jenny's bittorrent efforts have inflated that rate a bit. But I still need to work on the stability.

IP Binding works - I've got the Acer back onto his wonted IP addresses, after changing the address map, as enumerated yesterday. Tux still hasn't moved back to where he should be - he's on .65 at the moment, whereas I want him on .4. Which will involve amending a few scripts and browser bookmarks, to get the monitoring back up and running, once he moves - which may need a reboot.

Next question - should I instrument the other PC's here for snmp? It's probably overkill, but I'll bear it in mind.

And I need to get Emperor back up and running, which will involve changing his cooling fan, at least. The fan I now have, courtesy CPC, and I hope that it's all I'll need. Although I may need a replacement box, if the fan problems have resulted in more damage than I thought - maybe I'll need to buy another storage server. If I do, I'll name it either rockhopper or gentoo, following my policy of giving Linux boxen penguin-related names. Although, using gentoo as a server name could result in confusion with the distro of that name, so maybe feathers (McGraw, from Nick Park's "The Wrong Trousers" - GIYF)

Jane is away for a week, starting tomorrow. She's going to Weymouth, on the South coast. Sarah, at least, is going with her - as is Ben, the dog. I think Jenny is going as well, which will mean I'll have a peaceful week. But I'm on nights - there's always something...

Back to Daynotes


Saturday, 25th August

With the introduction of the new Homepages server, Demon provide some graphs and statistics for accesses to these pages - and it would seem that I have about 100 readers, world-wide. Thank you all for your interest in my semi-regular maunderings. I will attempt to provide more of the same in the future. Of course, if you feel moved to actually comment, I welcome e-mail - see the feedback page for details. But please, put a meaningful subject on any e-mail you send - the feedback addresses have been harvested by spambots, and although Demon's Cloudmark spamfiltering is doing a sterling job, some unwanted material does leak through. I cull material that does escape both Cloudmark and my own filtering quite ruthlessly - blank or gibberish subject lines are prime determinants for routing to the bit-bucket, unread, as are obvious pharma offers.

Jane and Sarah got away about lunchtime, with Ben the dog, but without Jenny - she's staying here. Apparently she wants to earn some money at her vacation job. She's also making noises about a gap year, while she decides exactly what she wants to do. Apparently, Physics isn't it.

But at least she appreciates that she'll get nowhere significant without a degree, and a good degree at that - grade inflation suggests that anyone without at least an Upper Second need not bother to apply these days.

So it's not going to be as peaceful here as I thought - but there's no dog!!! W00t! And Jenny is a good cook - as I type this she's cooking a steak pie. Her recipe is delicious, but it does require some beer - about 500ml or so. I ponied up a (UK) pint of Wells John Bull bitter, and got about a third of it back - that's a 4.8% alcohol by volume brew, and very nice too. The pie should go for several meals between the two of us, and since I'm on nights I will be taking frozen shop-bought ready meals to work - and maybe some home-made lasagna, again of Jenny's making. Yum to that, but the ready meals are only so-so.

Changing the router power supply has improved things - the time between reboots has lengthened to 8-to-10 hours, which is better than 1 hour or less, but still not good. I need to test the UPS by pulling the mains plug from it, and see what happens. And it would be extra-shiny good if I can get an idea of power consumption for all the core network kit. I have a power meter which can be put in series with any mains appliance - this would be a good time to try it. I misdoubt that the total power drain for Linkstations, switches, router and wi-fi is above 150 watts, but it would be nice to know.

Right, let's try it.

<pause>

After finding the device - it's a plug-in thing, with a UK mains socket on the front, and a 3 part LCD display - I pulled the mains input to the UPS, and interposed the wattmeter. The UPS grumbled at me while I did this, but everything kept running. So that's good.

And if the display isn't telling green back-lit lies, the core network bits are eating a measly 30 to 40 watts - that's the UPS charging, the Vigor router, the stairwell switch (Netgear FS108P PoE) one of the study switches (Netgear FS108) Tux the Linkstation HD-HLAN160 running Debian, and the Linksys WRT54GS V4 running DD-WRT. That's less than I expected, but in this case less is good, because all this runs 24/7.

And the ntp offset, as reported by mrtg is now recovering towards zero, after days of being all over the shop.

Back to Daynotes


Sunday, 26th August

Overnight Tux renewed his dhcp lease, and moved to the IP address where I want him to be - which, of course, fouled up all my monitoring bookmarks. It may also have confused some of the actual scripts - must check that, particularly ntp.

There were hard-coded addresses in 2 stanzas of mrtg.conf, both relating to Tux. I've changed them, and now I should start getting reports again. The actual programmes shouldn't have been affected - it's just their visibility to me.

But ntp seems to have got itself locked out of Demon's ntp server, the solution for which is a daemon restart. Do -

/etc/init.d/ntp-server restart

as root. This is for Debian - specifically Sarge (V3.1) Other distros may vary.

Mind you over the 10 hours or so that ntp wasn't talking to the world outside Semi-Detached Bedlam, Tux's clock only drifted about 1.9 milliseconds. This is better than 1 sec/day, not bad for a computer clock crystal.

And the router has stayed up for more than 24 hours. Long may it continue.

I've missed something - the access point isn't time-stamping in syslog. This will be a wrong IP address on the logging page there.

It was, but getting that changed involved an access point restart - so the Linksys is now counting up from zero again. Must check the data rates over wireless after this - maybe the burstiness will have gone. Or was it the router starting to go wrong? I'll probably never know...

Overtime tonight, a night shift. But fisrt I have to feed our neighbour's cat.

Last Week

Back to Daynotes

Next Week