Posts by Nigel 11 3191 publicly visible posts • joined 10 Jun 2009
The other thing you see is people who think that because the NAS box has twin redundant disk drives, there are two backups.
They really should make small NAS boxes with software that activates a red light and an irritatingly loud beeper when one of the drives fails. Because many people think they'll be magically informed when a drive fails, even if they've never told either the NAS box or the warranty-registration form their e-mail address, mobile number, etc.
Not that some of the cheaper ones would generate an alert even if you had.
Until someone developing a billion-item million-transaction-per-second system for HMRC gets paid the same as someone developing such a thing for Tesco, there will be a flow of people from HMRC to Tesco.
And follow this chain of thought.
HMG outsources to $ORG. $ORG needs some more staff. So it advertises with a salary sufficient to cause a flow of staff from Tesco to $ORG. $ORG then bills HMG for their services. Plus a profit margin. Plus administrative overheads.
And HMG now has to employ a contract manager to keep track of whether things are going OK. And a lawyer, when they aren't.
And this is supposed to save money??
I wonder how many of the people responsible for the outsourcing are now working for $ORG?
EFI is part of the BIOS. It is intended to be accessed as a filesystem. It's therefore bad design if corrupting that filesystem results in the Motherboard being bricked, rather than offering a "reset to factory" option.
But rm -rf is not the right way to erase a disk on Linux. Assuming that you are not unduly concerned to erase data beyond the abiliity of a three-letter-agency to recover it (*), then the right tools are
boot something like SystemRescueCD off a standalone CD or USB stick
DISK=/dev/sda # or whatever
smartctl -i $DISK # double-check you are about to nuke the right device
dd if=/dev/zero of=/dev/$DISK bs=4M; sync #write all sectors to zero
This also forces relocation of any bad blocks at a time when that will cause absolutely no harm, and lets you then use smartctl -A to see whether pre-emptive replacement of the disk might be wise. If you are in a hurry add count=10 to wipe only the partition table and anything lurking at the top of the first partition.
(*) if you are worried about the three-letter agency the recipe is remove disk, drill several holes in its HDA, smash its electronics with a hammer, and immerse the disk in acid. Coca-Cola is probably an adequate substitute for hydrochloric acid or ferric chloride. Easy to obtain and no bothersome H&S forms needed. Solid-state disks probably require incineration though applying a Dremmel coarse grinding tool to all the big chips might suffice.
Why is it that if you stay with a service provider for a long time, your costs slowly increase while your benefits slowly erode. Getting a quote from a new provider, using exactly the same criteria, will nearly always get much better deals
In tech, because the state of the art advances and the cost for service X goes down. But there's no incentive to pass on the new deal to an existing customer who just renews the old one by direct debit. Benefits erode? I've never noticed that.
In insurance, just because they can boost your premium every year if you just renew whatever they tell you. Bastards.
I have always found that the best approach is to use a price comparison site and then call your current service provider and tell them that you are happy to renew a decent service, but only at the price you found on the comparison site. That price may be their own, or that of a competitor offering (as far as you are concerned) the same service. They usually agree. At the end of the day be clear in your own mind how much extra you'll pay for the devil you know, if they won't price-match.
Last year it failed for my car insurance for the first time in about 20 years. So I changed insurer. The old one wanted three times what the competition did, and wouldn't budge, unlike in the past. I guess they just didn't want me at all.
What they don't seem to get is that customers talk to their friends. Most especially they grouse to their friends. If you treat their customers like crap there will be lots of formerly potential customers who will now shop elsewhere. And you won't know that it's happening. They'll say in private what they don't dare post in writing in public. Eventually you'll run out of naive victims to abuse.
By the way, every initially negative customer experience can be turned into an overwhelmingly positive customer experience. If something has gone wrong, apologize by deeds not just words. Put it right as efficiently and promptly as possible! You have one and only one chance at this - use it!
White goods. There's a smallish company out there that does it right. Starting with proper specifications easy to find (L,W,D of a fridge that had to go in a triangular space under stairs). Also customer reviews that I felt I could trust, including the product's "warts". Then a straightforward ordering process. Then delivery in a two-hour slot mutually agreed. Finally delivery drivers with mobile phones who are allowed to use them. (Mine called ahead to say he was running early, would that be a problem? )
Mark's Electricals. I think they're Midlands only. Nationwide soon, if there's any justice.
Recently Worktop Express pretty much replicated that experience, with online specification of every detail of a custom worktop in place of easy to find specifications.
Screwfix deserves a mention for trade/DIY supplies by click and collect. The system works beautifully. It has to -- if they don't get repeat business they'll fail.
There really isn't that much to it. Make it easy to find what you want and to order it. Let the customer know when it might be delivered before he orders (if there's a long wait then he may prefer to order something else). Give the customer a short timed delivery slot, and keep him informed on the day. I'm perfectly prepared to accept that traffic jams etc might mean keeping to the arranged time is impossible - so phone me as soon as you know of the problem to rearrange the time, OK?
What's the difference?
"The XYZ app's gone down" versus "The XYZ VM's gone down"
"Relaunch it!" versus "Reboot it!"
"A remote hacker could have executed his own code in the context of the XYZ account on the server because of a bug in the XYZ app" versus "... in the XYZ VM ..."
"Someone's rooted the server via XYZ" versus "Someone's broken in to the hypervisor via XYZ".
I think the main advantage of the VM is that you can migrate it from one CPU to another, even halfway around the world given enough bandwidth, without having to take it down at all. And that unlike the coding that would let you accomplish the same with an App running on a multi-site VMScluster, moving a VM is a purely operational matter that you don't have to code for.
(OTOH, when TS hits TF big-time I'd expect the VMScluster to come out on top).
The procedure "normally" was to yank out the de-commissioned wire before wiring its replacement, but after the build up you couldn't do it any more.
Cut the connector off both ends. It's easy to pull a dead wire through a bundle if it doesn't have a plug on the end. Not absolutely sure you are cutting the right wire? Get a cable tracer (above). Cables tied? Cut off the nylon ties and replace with velcro ties, one at a time.
OK I have never tried any of this on a three-foot bundle. In theory it works.
Guys, buy yourselves a cable tracer! (aka "Tone and Probe" kit)
Plug the signal generator thingy in at one end, Wave the probe at the patch panel outlets or the cables protruding from them until you hear the warbling tone loud and clear. That's the other end.
With 100BaseT you can even trace a cable without breaking the datacomms for more than a few seconds, You inject the tone on the 7/8 and 4/5 pairs which 100BaseT does not use. 1000BaseT uses all eight wires, but much hardware will downgrade itself to 100BaseT if you disconnect 4/5/7/8, and most users won't notice the slowdown). You use one of those 2:1 network splitters as the injector. (You know, one of those thingles that connects 4/5 to 1/2 and 7/8 to 3/6 on the second outlet so you can run two 100BaseT Ethernet channels down a single Cat5e premises cable. Naughty but occasionally nice, compared to a two-week wait and a three-figure bill to install a printer).
Why? Well, there are nightmare undocumented spaghetti wiring tangles. And then there are anally retentive tightly bundled and tied installations that look like works of art, except that someone put the wrong numbers on the ends on one or more wires. Both problems easily solved.
Yes, you can have fractional orders of magnitude. Its just use of logarithms. 10 is one order of magnitude ( log_base_10 of 10 is one). Half an order or magnitude is antilog_base_10 of 1/2, which is of course equal to the square root of 10 by definition.
sqrt(10) is 3.1622... but since orders of magnitude usually imply low precision handwaving, half an order or magnitude is a factor of three. I've never heard anyone use "a third of an order of magnitude", but the cube root of ten is quite close to two.
OUt of interest - do you get a few years life out of a car battery? I've always understood that the gel-filled batteries in UPSs are deep-discharge-tolerant, whereas car batteries are not. So if you run a few times on the battery until the UPS decides it is time to give up, with a car battery you may find that it is knackered. But of course, three-hour power cuts are pretty rare.
The mains is much more reliable than your average UPS.
Not even in West-end London, if the UPS is an APC SmartUPS. That's true even if I exclude the various times it was a premises fault(*) rather than a fault with the supply to the premises(**).
Now I'm out in rural parts, I see power outages or glitches several times per month. Sometimes the desktop clients keep going while the lights flicker, sometimes they crash, sometimes the lights go out for a second or two. Always, the UPS has kept the servers going as if nothing happened.
(*) e.g. I once plugged in a faulty PC to diagnose its problem, and the PSU went bang, and the lights in my office went out. Trudging down the corridor to the circuit breakers I couldn't help noticing that everyone else's lights were also out! The outage was traced to a huge fuse in a box high on the wall above the breaker panel that had probably been installed in 1920 and not maintained since. The electricians sourced a replacement in Manchester and paid a man on a motorcycle to bring it to us asap. Possibly that was the last fuse that big in the country, or even in the world. The alternative would have been a whole of site shutdown and major-league rewiring of the building. Oh, and the UPSs kept the local servers up for ten minutes and then initiated a controlled shutdown when their batteries were running low.
(**) Google "exploding pavement".
Argh. You can't just post that without giving out the juicy details
To make up for not knowing those details, here's my small rack-mount UPS story.
Somebody ignored the UPS replace battery warnings, and the battery pack swelled up a lot. It forced the sides of the UPS case outwards a few millimeters which was enough to make it completely impossible to slide the UPS out of the rack on its runners, and also completely impossible to slide the very important server immediately above the UPS out of the rack.
The idea of leaving the UPS there "forever", turned off and disconnected, until the entire rack was scrapped, was given some consideration. But the thought of a high-pressure acid leak in the vicinity of much expensive hardware was too disconcerting. As was the (lack of) gravitational stability of a rack with many kg of UPS near the top and the chances of doing one's back an injury in getting it there.
It took several hours downtime. We had to poweroff everything in the rack. Then we had to remove servers one at a time, starting at the top, until the one immediately above the UPS was able to move a few millimeters upwards and unbind itself from the body of the UPS. Then there was lots of poking and twisting and prodding with various tools down the very restricted spaces between the UPS and the rack until eventually, it was made possible to extract it using brute force. Then we had to reassemble everything in the rack, this time with a 1U gap above the replacement UPS "just in case".
No, the old UPS didn't explode. Not even when a colleague was kicking it hard from behind to force it forwards a few millimeters at a time, while others held on to the rack.
Argh. You can't just post that without giving out the juicy details!
I don't know the details. Just that the fire alarm went off, that after we got back into the building the servers in the datacentre were down, and that people I know were in charge of the things in there were shouting into phones and running around like headless chickens. Later I found out that the company doing the rolling maintenance on the big UPS got it seriously wrong.
Basically, the UPS went KZERRT and a tonne or so of lead-acid batteries makes for a really loud KZERRT, including smoke. Somehow they managed not to dump the Halon, and nobody was hospitalized (neither by the UPS nor by the management folks).
Just read and follow the instructions which come with an APC SmartUPS 750, or every previous APC model that I have dealt with. There's a small quiet KZZRT when you connect the new battery, but you are warned to expect it and it's harmless. The wiring you are playing with is (a) low battery voltage and (b) well-insulated, unless you are the sort who feels an irresistible urge to stick a screwdriver into a battery connector.
The first time I did his I lacked confidence, so I arranged a scheduled shutdown of the server and substituted a 60W lamp for the server while I swapped the battery. Not a flicker.
This is one reason why I am an APC fan (although the company has been taken over since I last purchased one). The other is their longstanding open-source support of their Smart UPS protocol.
You do need to consider the slight risk of a power outage while the UPS has no battery. You have to consider "if anything can go wrong, it will", and "we need a scapegoat". Large datacenter-sized UPS battery banks are replaced on a rolling maintenance basis by the support company. I was there once when that went horribly wrong. I prefer the small stuff myself.
I should think that there are some large corporates with VMS cluster incarnations that have existed since shortly after VAXclusters went on sale. (From dim memory, circa 1990). This is of course a "Grandfather's Axe" entity. You can replace every part of a VAXcluster (now VMSCluster) many times over without ever requiring a new incarnation. Given enough bandwidth between sites you can also relocate a VMSCluster datacenter without having to create a new cluster incarnation.
Anyone care to post their VMScluster incarnation date?
As far as long-running single computers are concerned, Voyager 1 will take a lot of beating. Launched in 1977 and still working.
Where do these people get an eighteen year supply of uninterrupted electricity in broom cupboards from?
From a (smallish cheapish shoebox-size) UPS?
If you live in rural parts short power outages are frequent and so are voltage excursions (sags and spikes). If you connect a PC direct to the mains, its power supply is likely to fail within a few years, quite apart fron whatever the crashes will do to its software and filesystem integrity. If you want a quiet life you insert a UPS between the PC and the mains. In my experience a shoebox sized APC UPS will protect the PC from all of these and won't be taken out by anything short of a nearby lightning strike (the sort which also fries most of your domestic electronic appliances).
Every few years it will tell you to replace the battery, which you do "live" without taking the PC down. (*Do* replace the battery when told to -if you don't it will swell up so you cannot get it out, and some time after that it may "explode": spray sulphuric acid all over its vicinity. )
And incidentally, the one I have will keep a small idle PC running for about half an hour, so only the most serious of power outages will take your PC down. So yes, I can believe 18 years. The last outage in excess of half an hour was the 1987 "hurricane", although of course, others haven't been so lucky in more recent years. And no, I'm not claiming uptime since 1987.
There should be another application; the glass in solar PV panels.
They need to self-clean as well as possible, so dirt does not block light from the panel. They also need to stay cool, because efficiency drops as they heat up.
The best place to put them is a desert, but deserts are typically hot, dry and dusty, and water in a desert is an expensive commodity. Desert panels need to be washed, and if this glass means they could use less water to wash them just as well, that's a big advantage.
it might just take the rest of the evening to get back to anywhere near desirable temperatures
What do you live in? A mediaeval castle? Or are your radiators and boiler woefully under-specified? (Or you might find that your radiators are full of sludge and emitting heat at less than a third of their rated output).
I've always programmed the timer to turn the boiler off at the time I leave for work and on again about half an hour before I'm expecting to get back. Plenty warm enough when I get in. This in a Victorian solid-walls flat with draughty sash windows (albeit with the assistance of waste heat from downstairs who are always in)
Being able to turn it on using my cellphone, or even better to program my cellphone to turn on the heating when the GPS tells it that I have started my return journey, would save me significant amounts of money and CO2 emission.
haven't ever seen a Tado but ... are you sure that your car hasn't worked out that if it jumps up the wall and bashes the thermostat often enough, the thermostat thinks someone is still in the house and turns the boiler back on? My parents' cat learned to ring the doorbell to be let back in. There was a cat-flap, but the doorbell was more fun. Another cat learned to open the fridge to steal food and then to close the door behind it, presumably because it did not like the draft of cold air coming out of the white box.
Cats are the opposite of Artificial Stupidity: well-honed Natural Intelligence, adapted to running on a smallish CPU with far too little RAM. (If they always rembered what they worked out yesterday, we wouldn't stand a chance).
Yes, it does seem like an ideal Raspberry Pi / Python project.
The one thing I have not found yet is something like Honeywell's proprietary remote-controllable thermostatic radiator valve heads. Anyone know of an open equivalent (or how to hack the Honeywell ones? ) I really don't fancy installing zone valves and computer-controlled mains wiring on every radiator, the c/h plumbing is buried in the walls, and treating the whole house as a single heating zone is pretty wasteful.
Not years. Decades of uptime. It was the individual nodes that clocked up years. You could migrate the fscking architecture (VAX to Alpha) without requiring a new cluster incarnation. You could relocate the datacenter without requiring a new cluster incarnation (though that trick did require a really high bandwidth fiber link).
Oh, what might have been, if Digital hadn't been subverted by some idiot PHBs (running Microsoft indoctrination OSes in their crania, or possibly their colons).
I have to hand a Compaq 80286 system that still runs Windows 3.1 perfectly on its original hardware. To be fair, it's not a server, and has spent a good fraction of its life switched off. It was the control system for an expensive piece of lab apparatus that was (relatively) recently scrapped because it broke beyond our in-house abilities to mend it, and the manufacturer went bust many years ago.
I don't have the heart to throw it away. Will probably fire it up annually at Xmas party time until it won't boot, or until I pop my clogs, or until the National Museum of Computing or NASA wants it.
It has a 5.25 inch floppy drive.
Linux will soon breakthrough from irrelevance to rounding error in the stats.
Ever heard of exponential growth? Not saying we've got it, but if you mean that we're grown by an order of magnitude, then world domination may be closer than you think.
At the very least we'll be ready for a big push when Microsoft follows Digital into self-inflicted oblivion.
It took a second to download and a few more seconds to install.
The Linux upgrade process these days is in a universe Microsoft hasn't even thought about visiting.
One should perhaps also mention that with a little forethought, you can snapshot your filesystem prior to applying the updates, and if you do not like the results you can revert to the snapshot in next to no time. This includes in the very unlikely event that it won't reboot - you just do the reversion from a standalone CD instead.
What do Microsoft folks do? How long does it take to "revert"? And do you get your disk back to the exact same state it was in prior to the failed updates?
This is beginning to look like an offence under Section 3 of the Computer Misuse Act, 1990.
A plan, for anyone whose applications software is dependant on Windows 7.
Send a letter via recorded delivery to MS UK's registered company address (address for serving formal legal papers), stating that you are running Windows 7 and that you require your systems to continue running Windows 7 and that you will seek a prosecution under the above act should they cause your computers to cease to run Windows 7. Better still, pay your solicitor to draft and deliver this letter.
This might make them think harder. It will certainly help you no end when you sue them for damages. They ignored the warning. They cannot claim that the damage to your systems was accidental.
Personally I'd prefer to see Microsoft hurl themselves off the cliff into a lawyer-infested pool. Why save them from the consequences of their own folly and arrogance?
The business I work for uses Linux for almost everything. The exception is Autocad, where Microsoft has purchased a monopoly for itself. It's already running on Windows 10. The users detest it, but they have learnt how to fire up Autocad and that's all that really matters to the biz.
Can you imagine? After a long weekend, come to work to find Win 10 has installed itself on most of a company's PCs, and 'only a few' applications have been disabled in the process. Like all the key ones, such as their ERP system.
Or the embedded control systems for their £100K+ production machines, which are the sole reason for the company's existence, and without which they will be out of business in a week? For which a Windows-10 compatible version of the software doesn't exist, because the manufacturer of the machinery went out of business a few years ago?
I was also going to offer MegaBank's cash dispenser network, but the folks who bought you the banking crisis and who manage to have mainframe FUBARs several times each year cannot be that incompetent. Can they?
@ stucs201 No, and I suspect high CRI G9 led isn't do-able for lack of any place to sink the extra heat that filtering out the surplus blue components will generate.
LED G9 is pretty marginal anyway. I experimented with "ordinary" G9 capsules and both failed within months. I took them back for a refund. Luckily I had kept the receipt, knowing that LEDs need a heatsink and that there is no space for a proper heatsink in a G9 capsule.
It's sometimes more cost-effective to replace the light fitting than to retro-fit LEDs. Or at least it is if you are your own electrician.
People may be willing to pay a higher total lifetime cost to save the extra CO2 emissions from the Halogen filament bulb, or to get a CRI even better than the best LED (around 97 at present). Also they said it has potential to be 30% efficient in converting electricity into visible light whereas the best LEDs judged on light quality / CRI are around 8% (*). So that's the lifetime electricity cost down to a half or even a quarter of the LED.
But this is research, nowhere near product development yet. First question is whether they can make them in production quantities at all!
(*) The crappy bluish ones are around 13%, because one of the ways to make better quality LED light is to filter out the surplus blue component and turn it into waste heat.
A good fraction of the heat generated up near the ceiling goes up through that ceiling. If you live in a flat you are making a donation to upstairs' heating bill. If you aren't using the bedroom above your living room until several hours later, that's largely waste heat in winter and an even more uncomfortably hot bedroom in summer.
Also per BTU or KWh or Joule of warmth in your house, Electric costs more than gas (and oil at today's price!) so it might not be a waste but it's still an extra drain on your wallet.
Anyone who thinks LED bulbs are little better than CFL bulbs is buying the cheap ones with a CRI (Colour-rendering index) of 80. They range from no better than a CFL (except blue rather than green) to just about OK. If you have to buy in a shop in the UK, Philips ones seem to be the best of a bad bunch.
You get what you pay for. (I.e. if it's cheap it will not be good, although if it costs a packet it won't necessarily be any good either).
Search out LED bulbs with a CRI value of 90 or higher. They are remarkably hard to find in the UK where I guess they know that most folks look at the price not the specifications. The best LEDs I have found are made by Civilight and have a CRI of 95 or higher. They're dimmable. The 2700K warm white ones are sufficiently close to Halogen that I'm perfectly happy to spend £15 each. (I had to get an energy assessor to come back because he took one look and decided they were halogens! )
There are a number of other less dear brands available in Germany boasting CRI of 90 or above. In Germany people seem to be more willing to pay more for a better product.
How to get them here? I've found that the easiest way is to shop at elv.de using Chrome and its built-in Google translation option. ELV stocks a large range and all the specifications are there. The hardest bit is feeding a UK address into a German shipping form. The EU means there are no customs or other hassles, shipping is not particularly dear, and the biggest rip-off is the extra percentage your credit card will charge for Pounds to Euros conversion.
Oh, and pay attention to the colour temperature as well! I assumed that because the 7W Civilight bulbs were 2700K that the 10W ones "warm" ones would be likewise. They were 2800K, rather whiter than I wanted. I can see why people hanker after the days when a light-bulb was just a light-bulb.
my experience is that if you fully format a pen drive you have something like a 10% chance of permanently corrupting it, making it unreadable
In which case you have identified a crappy unreliable pen drive before storing anything you wanted to keep on it, saving yourself much grief later. Buy a different make next time. And remember Sturgeon's law.
An interviewer asked him what happened to 'live fast, die young' and Lemmy replied, 'I tried, I really tired
Which is why naming an element with an extremely short half-life after him would be so apposite. The atoms are trying, really trying, but it's their half-life we measure and a few always get lucky.
Until there's only one left. And then, none.
Someone also needs to comment on the "live fast, die young" ethos which I think was embraced by the man and is certainly embraced by the element. Note that both for elements and men, that attitude doesn't guarantee an early demise. It just means they are rolling the dice more often.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
