Hacker News Reader

Comment by foobarian | Hacker News Reader

I for one do not mind that part of computing getting abstracted away by the cloud, whatever its other failings. :-)

Working on the hardware directly is more than half of the fun.

An old IBM tech told me that hard drives failed when power cycling equipment could be temporarily resuscitated by placing the drive on the floor, twisting it hard by hand, then as quickly as possible connecting it back into the machine and powering on. Apparently getting the platters to spin helped them get around a worn out motor that couldn't make the initial spin-up.

I remember upgrading from a 4.3 GB HD to a 20 GB HD around 2001. I had the old hard drive sitting on my desk, and the new hard drive sitting in the case, both attached to the motherboard via IDE ribbon cables.

Most of the way through backing up the 4.3 GB HD to the 20 GB HD, I heard a screech and the old hard drive twisted slightly on the desk... conservation of linear momentum. The drive was visible to the BIOS, but refused to spin up after that.

I put the hard drive in a ziplock bag, put it in the freezer to give the parts slightly more clearance from thermal contraction, and slammed the hard drive nice and hard on the desktop to free the stiction. That revived the drive long enough to finish my copy.

I did have backups of all of the really important information, but restoring from a stack of floppies is more tedious, slower, and less fun than percussive maintenance on a hard drive.

Back when I had an internship working on MEMS gyros for GPS guided mortar rounds, we had to sometimes perform similar percussive maintenance if static electricity had caused the moving parts to contact the substrate. We took the gyro, and smacked it hard on the desk in an "eyeballs out" orientation to give it something like 10 to 100 Gs of acceleration in an attempt to un-stick the MEMS gyro.

Yeah, did same in 1999 or 2000, but we _did_ _not_ take the HDD from the freezer, just pulled the ribbon and power cable through the narrow slit between the door's rubber gasket and the fridge's body.

Clearly, I meant conservation of rotational momentum, not conservation of linear momentum.

Yes, stiction is a real issue in old hard drives. In fact something that used to work well if you were cold booting an old server even in the mid 90s, with the expectations of an immediate migration, was giving the drives a good sharp slap to help break stiction so they could spin up. The only issue is it could cause a head crash. Another fun trick from back in the day was sticking hard drives in the freezer to cause the platters to shrink enough to dislodge the head and restore the air gap if you had a head crash so you could try to recover the data.

As much as I love hardware, I much prefer our current solid state wonderland.

> to shrink enough to dislodge the head and restore the air gap

Huh?

What I heard is what a cold drive would be a bit more magnetically stable and IMMSMV that surely worked, because I never had a drive with a stuck heads but I did had drives what would just abort the read or return gibberish, but after an hour in the freezer they would read just fine, till they heat up again. That's for the drives mfg after 2000, if that matters.

My understanding may be incorrect, after all the freezing the drives trick is mostly something that was shared among sysadmins like an old wive's tale. Nonetheless, it does work (or did).

The way I had always understood it was that if you had a head crash (which was caused by the head physically contacting the platter, overcoming the resistance of the air gap between the head and platter, usually due to physical impact) that the magnetism of the head would prevent it from lifting back up on its own, and that freezing the drive would cause the metal in the platter to contract away from the head, which was restricted in movement by its armature, thereby restoring the air gap and lifting the head away. If you started the drive spinning before it heated, the head would stay out of contact and you could successfully read data (some of it, for awhile).

That said, my understanding may be incorrect.

I would believe them. I have heard a similar story. A data center had to go through a planned power cycle event.

All systems were working when they shut down, but ~40% of the disks failed to spin up due to worn down motors.

That's a lot of disks.

What’s a procedure to avoid it? Reboot machine once a week so disks fail not all at once?

Gotta know which way to twist it.

> Working on the hardware directly is more than half of the fun.

It never really worked until you gave blood to the computing gods on a sharp edge of a metal case.

I feel like the general improvement in fit & finish of cases and chassis is overlooked.

I, for one, haven't flayed open a knuckle on a random razor edge or burr on the inside of a case in like 15 years.

"Hey, look, I have all five fingers! Three on one hand and two on the other!"

I have 6 stitches in my left hand from such an offering.

Unless it was a recent incident, you should probably see a doctor about getting those stitches removed.

Looks to hands and fingers intently

Hmm, yes, true.

Every Friday I had to take a bunch of magnetic tapes to the bank so they could look after our backups ( which was a good excuse to pop into the pub next door to the bank...) Oh, and we never ever tested that the backups actually worked

My first job out of school was engineering consulting, writing software models of pre-production military radio hardware. The client was a military contractor, and they paid a crazy sum in order to have 3 DAT tapes: one tape always in the drive used to make the daily backups of the shared working directory, one tape always stored in a bomb-proof bunker, and one tape potentially in-transit via courier. I think it was once a month that the client's sysadmins rotated in the "unused" tape, sent the latest tape to the bunker, and waited for the old tape from the bunker to return as the "unused" tape.

Due to absurd government requirements, we were not allowed to use any version control software other than PVCS, and PVCS could only be used for storing official releases of the models. So, the last person in the office on Friday needed to make a dated zipfile of the shared project working directory.

So, one day I convinced the project manager to just let me rename our working directory, create a new empty one in its place, and tell our sysadmins that I accidentally deleted the contents of our working directory. It turns out that the client's sysadmins hadn't set up anything to expunge the oldest backup and had dutifully ignored the alarms that the tape was full. So, the latest available backup was from about a week after they last changed tapes!

The senior sysadmin prominently had a sign on his cubicle reading "Programmers are the ditch diggers of the 21st century." Those same sysadmins were supposed to keep the server room door open while I reinstalled Solaris on a Sunfire V1280 via serial connection from my laptop, because there were cables carrying TSSCI (Top Secret - Secret Compartmentalized Information) traffic. However, the V1280 consumed so much power that the server-room was under-cooled, so the sysadmins illegally shut the door to keep their cubicles cooler. They had top-secret clearances. They knew their obligations to keep me away from those TSSCI cables. Those lazy lazy arrogant sysadmins.

Anyway, unless you have recently passed a restore-from-backup drill, you don't really have backups. Even if you pay absurd amounts of money for sysadmins with top-secret clearances and absurd amounts of money to store your backups in a bomb-proof bunker.

Reply on news.ycombinator.com