28th February

It was pointed out to me today that I needn't do all that extra work on the server cabinets, as I could just fuss the supplier into providing the proper rails, or at least keep a better eye on my manager to make sure that he orders the right options... It's a fair point, but I think the main reason that I don't is simply from a long-established state of mind: I started work in the IT industry in the early eighties, when PC networks were just starting to be taken seriously but the mainframe was still king. As a result, budgets were low and management expectations were high, and this meant that improvisation, bodging, and out-of-spec modifications or upgrades were very much the order of the day... then my next few jobs were at small companies with a miniscule budget for IT, and this strongly reinforced the shrug, roll up your sleeves, and make it work attitude.

The PC hardware itself was in it's infancy, too, with almost infinite variations on internal design details... I can't guess at the number of times that I installed a second hard disk in some nameless clone by wedging it in with anti-static foam, rolled and taped into a tight cylinder, because no suitable drive rails were available... The big-name systems were often little better, with each new delivery bringing subtle changes in the fine details - screw hole spacing, drive bay height, the weird non-standard connectors...

Finding adequate sales or technical support for issues like that could be almost impossible back then, too - most of the UK suppliers were just re-selling tweaked-over US or Far Eastern imports, and trying to persuade them to ship you something as unglamorous as drive rails could be an exercise in futility... And in the pre-Internet days, with only a minority of manufacturers having BBS systems, I remember once waiting at least a month for a network card driver disk to be shipped from Holland.

So when I'm faced with a server rail that won't fit into a cabinet, my natural thought really is to bodge it to fit - and usually it's only later, as I'm mid-way through attacking it with hacksaw and pliers, that I think of the warranty... Maybe it's a bad habit, but my ability to get a system running using only a short length of string and a double-ended screwdriver has saved the bacon of many managers and users over the years, as well as my own on more occasions than I care to count.

Oh, and the server rails fit very nicely this morning, thank you very much...

27th February

There are holes in the sky,
Where the rain gets in.
They're not very large,
So the rain is thin

    Spike Milligan, 1918 - 2002

He was very, very old...

We came to mount a test server in a new cabinet, today, and discovered that the frame uprights couldn't be moved far back enough to accommodate the server rails. I brought them home tonight to work on, and while I was machining the bolt slots out a little longer with the Dremel [Attempts Tim Taylor grunt] I started musing on how many times I've had to modify server cabinets over the past few years. The smoothest ride came from Compaq servers in a Compaq-branded cabinet, but even then I had to hacksaw a couple of the little cosmetic blanking strips on the front rail... and nothing fit properly in the Metastor cabinet, except the Metastor hardware, requiring many spacing washers and long bolts. The recent batch have been Prisms of various sizes and depths, and all have required modifications ranging from drilling out bolt slots, as today, to bending parts of mounting rails over the computer room doorstep... Part of the problem is that the server manufacturers are trying to make their mounting rails more elegant - the rails for the current crop of Dell PowerEdge systems look like they would just click into a subtly different (Dell approved) cabinet, but on the Prism's rails the cleverness just doesn't fit, so a little metal tab has to be bent ninety degrees with pliers.

I once watched the official Compaq "How to install your rack-mounted server" video, and it never mentioned anything like this... It's a pain, usually, when I'm on a deadline and surprised by one of these awkward misfits, but somehow it's comforting, too - even in the 21st Century, network admins still sometimes have to cut steel with fire.

26th February

Panasonic's next-generation SuperDisk is finally available in Europe, bringing not only the promised doubling in capacity from 120Mb to 240Mb on the special disks, but also a remarkable 32Mb squished onto a regular, bog-standard 3.5" floppy. Both options seem fairly attractive, although I have to say that the LS120 drives we have at the office have been a touch cranky at times... Nothing one could raise a support call about, but just occasional odd behaviour and a very occasional BSOD. I think most modern removable media is like this, though - my little ThumbDrive flash disk has certainly had it's moments, too. It was great for the first few months, but then seemed to become crankier and crankier, to the point where (even after much cleaning of old drivers and careful re-installation) I won't bring it near my desktop for fear of blue screens... I noticed myself sniffing at external USB hard drive enclosures today, though, and really must have another try with the ThumbDrive before spending any more money on portable storage...

Apropos of nothing, I really want to mention the Australian tech site Dan's Data. Dan's reviews (of everything from PC hardware to radio-controlled tanks to liquid magnets) are full of accurate information, useful opinions and hard science - all written with a marvellous twist of geek-humour that really tickles me.

Try these articles for a fair cross-section of the three hundred and forty he has online:

Cooling your PC the SERIOUS way

The top eight computer myths!

Tabletop Trebuchet kit

How to spot a psychopath

25th February

The replacement KVM user station expired over the weekend, with the same symptom as the original... my working theory is that there is actually a fault in the KVM switch itself, which kills anything connected to that particular port... I've seen it with SCSI controllers, and see no reason why something with as much brain as that KVM switch shouldn't be behaving similarly. Raritan will investigate the two expired units, so we'll see...

Some odd news in The Register, today - firstly, an article by Thomas Greene containing a scathing attack on veteran IT guru Steve Gibson... Now, I know that Gibson is not universally admired, but I've been using his software since SpinRite in the days of MFM hard disks, and reading his opinions on TCP/IP security for almost as long - and I'd be extremely surprised if he is actually the arrogant, know-nothing bozo that Greene claims...

Secondly, and rather more positively - District Court Judge Marilyn Hall Patel, currently overseeing the case of Napster vs The World, has decided that the recording industry consortium is "attempting the near monopolization of the digital distribution market", and that "The resulting injury affects both Napster and the public interest." Well, personally I strongly agree - but Patel has apparently been heavily biased against Napster in all her previous statements, and this sudden U-turn is completely unexpected! Whatever next? Scott McNealy revealing that he's signed a deal with Microsoft to port Windows to the SPARC platform?

I've found something neat, too - the US firm Errorwear are selling what appear to be most excellent T-shirts, emblazoned with the "Blue Screen Of Death" errors from various versions of Windows. They also have designs with fatal errors from Commodore and Apple platforms, as well as the HTTP 404 Not Found text from both IIS and Apache web servers! Unfortunately the only black shirts, my usual choice of colour, are the Amiga's "Guru Meditation" and DOS's classic "Bad command or file name" - neither of which are quite my thing...

On the subject of DOS - I picked up one of my old-favourite SF stories last night, Bruce Bethke's "Headcrash" - and to my surprise noticed that when it was written, Bethke was working as a technical writer for my company's arch-rival. Small world! He seems to be working for shadow-of-their-former-self supercomputer company Cray now, though, and the fact that he wrote the novelisation of the stunningly unsuccessful movie "Wild Wild West" suggests that his writing career is heading into somewhat of a decline all round...

24th February

I finally nerved myself to disconnect the cable loom and lift the PC off the desk, then fitted the new drive cooler and some more sound insulation. The left-hand side panel doesn't have clearance for the "Magic Fleece", so I used the thinner cork-based sheets. They were far easier to work with, but it still hasn't made much difference to the noise... We think maybe a little... This means that either the noise I have isn't at the sort of frequencies that can be damped with this kind of material, or that it's escaping through areas that I haven't yet sealed - I shall try the front panel, next, which will be fiddly but worth doing for completeness if nothing else. At least the case temperatures don't seem to have increased significantly...

Old and new coolers - and the Fan Counter incremented by one... I think there are fifteen in INFINITY, now, more than half the entire network's population!

I've finished the Redstone kit, too, and it's looking rather better than I'd feared at various points - for such a simple little model, it's caused considerable head-scratching. Images on the Models page as soon as I can find some more batteries for the digicam - my old Kodak DC120 was a cutting-edge camera in it's day (sometime around 1870, it feels, in comparison to the current offerings) but is a real power hog...

23rd February

I've made extensive updates to the Space Models pages over the last 24 hours. A couple of them will be rather slow to download over a dialup connection, I'm afraid, but you can't break an omelette without making eggs...

It's been a day for oddments, today - some work on the decals for the Redstone (not that I've finished painting it to my satisfaction, yet!), most of the finishing touches on the Von Braun Lander, and some sound-proofing on the PC.

That Magic Fleece is seriously odd stuff: around a centimetre of dense felt-like fibres with some kind of viscous layer in the middle - almost impossible to cut with a knife, but scissors fared better. The instructions were all in German, but I assumed that they were the standard "clean surfaces before-hand, apply carefully to avoid air bubbles" sort of thing, and proceeded accordingly. I've plastered it over the right-hand side case panel, the easier of the two thanks to the absence of extra blowholes, and will wait to see how it settles in before proceeding with the rest. I have to admit that it hasn't made any apparent difference to the noise level, as yet...

[Later] Added a few spare decals to the lander, which I now declare finished,, and decalled the Redstone too while I was at it... even though I have a little more painting to do around the tailfins, yet! The lander decals were really odd - all sorts of white flakey stuff floating around in the water, possibly due to age - while the Redstone's managed to be both rather thick and annoyingly curly at once, as well not being very nicely designed... I eventually peeled a couple of them off again (boy, that was hard!), as they just didn't look right...

I'll try to finish the Redstone tomorrow, and then that's probably it for the space kits, at least for a while... I've built most of the ones that I like the look of, now, and Ros has been making hints about help with her doll's house as the next project... Working with wood instead of plastic will make a nice change, though, and doubtless I can use the Dremel for a million useful things. Watch this space...

22nd February

Not the best of weeks... I was testing the new open file management system, and all did not go according to plan. This really has been a Project of Lost Souls - we bought Legato's own open file add-on around a year ago, only to find that it would only run on the new (very new... too new...) version six of the Networker backup suite. I reluctantly upgraded, which went remarkably smoothly but brought few obvious improvements or new features, and after a short settling-in period tried the Open File Connection. There were immediate problems with backups failing and clients hanging but the subsequent support call revealed the startling news that OFC was already an end-of-life product, and that we should switch to their new offering, a licensed version of the industry-standard St Bernard Open File Manager, then at a nice mature version 7. This went extremely well in initial testing on the secondary servers, but proved disastrous when exposed to our main data store, a big-ass Compaq Proliant with MetaStor RAID, currently holding around 100Gb - and which also hosts Networker and the tape library...

Not to put too fine a point on it, the server went repeatedly and completely tits-up shortly after starting any significant backup operations, and sometimes smack-bang in the middle of the day just for the hell of it. It didn't meet any of the classic criteria for OFM to behave like this; there's no anti-virus software installed, Diskeeper is safely out of the backup window, and even throwing another 256Mb of RAM into it on general principal didn't help one little bit.

"Not to worry", said Legato - "Version Eight is shipping soon, and that will solve the problem". It took a while to find out that it wouldn't, as it took around five months to get a working serial number. Many things were tried, and many, many license codes and enablers were issued by our supplier, by Legato, and by St Bernard themselves...  eventually I got to the point where I was entering each new code directly into the test server's registry rather than having to uninstall and re-install each time... Last week the bureaucracy finally spat out a working enabler, and I sprayed it liberally around the secondary servers.

A few tweaks were needed, but all appeared to be working well... Nevertheless, I had a sense of foreboding at the thought of letting it near the big iron again, and had to nerve myself to perform the (admittedly rather neat) remote service installation on Wednesday... That night's backup wasn't very successful - the dozen or so secondary servers streamed through beautifully, but when it came to backing up itself, the server hmmmed and haaaaahed and eventually decided to time out. I wasn't too happy with that, but tweaked a few more settings to account for the larger file-system - and then Thursday night's run seemed to go well.

As usual, recently, the Exchange backup group was stalled when I got in on Friday morning, waiting for a tape from whatever random pool it had decided it just absolutely must write to, today... I gave it a tape, and it was still churning through on low priority at around eleven, when NT decided that it no longer had any free resources and dropped all it's protocol stacks, one after the other, and then hung the shell.

So - it's the middle of the day, there are probably a thousand files open, a backup is running, and the RPC handlers are down. Ouch. Big Red Switch time.

And then it blue-screened when I tried to uninstall the OFM, ten minutes later.

The air was blue, too.

But it's Ok, now, and I've told my PFY that if he ever sees me heading towards that server with anything even remotely resembling open-file software, he is to wrestle me to the ground and soak me down with the Thorazine spray.

Enough said.



Click here for 12th Feb to 21st Feb



Weblog Archive