Earthquake Country: Your NAS Is on a Shelf and Your "Offsite" Backup Is Eight Miles Away

Almost every disaster we have written about on this site arrives through something. A surge comes up the power lead. Water comes down from a ceiling or across a garage floor. A ransomware infection comes in over the network. In each of those the damage happens to the electronics, and the useful advice is about wires, plumbing and software.

An earthquake is not like that, and the difference matters more than it sounds. It does not touch your data at all. It spends ten or fifteen seconds moving objects that were not attached to anything, and then it stops. Your files are not corrupted. Your backup is not encrypted. The drive is simply on the floor now, and so is the network box that used to sit above it, and the bookcase they were both standing on is lying across the desk.

There is a date on this one, which is unusual and worth using. The Great California ShakeOut — the statewide drill — is Thursday, October 15, 2026, traditionally at 10:15 in the morning. The California Department of Education wrote to county and district superintendents on August 28 to tell them so, noting that most schools will drill on that day and that last year more than 10.4 million Californians took part, including more than 5.4 million K–12 students and staff across more than 450 school districts, 320 charter schools and 460 private schools.

So there is a deadline, it is about six weeks out, and it applies to the whole state rather than one neighbourhood. What there is not is anything written for you. Search for how to protect your computers before a quake and you will get three kinds of page: insurance and continuity PDFs written for companies with a risk officer, vendor blogs for cloud backup services that would like to sell you the answer, and university safety pages written for a campus. The advice in them is real but it is scaled wrong — it talks about anchoring server racks, restraint systems with elastomer bumpers, and contracting with a colocation facility in another part of the country. Nobody has written the version for a house in the Valley with a desktop, a network drive and a printer, or for a shop on a boulevard with a register and a machine in the back office.

That is what this is. It is the physical half of the job, which is the half almost nobody does, and it is cheap and it is fast.

The failure mode is mechanical, so the fix is mechanical

It is worth being precise about what you are actually protecting against, because it changes what you should spend the afternoon doing.

A power surge destroys electronics from the inside, through the power supply. Water destroys them slowly, by corrosion, over the days after it dries. Both of those are invisible while they are happening and both are best handled with equipment — a decent surge strip, a battery backup, getting things off the floor.

Shaking does none of that. It applies a sideways force to objects for a few seconds. Anything heavy and unattached slides, and if it slides far enough it falls. Anything tall and unattached tips. Anything on a shelf above head height becomes a projectile. The electronics are perfectly healthy the entire time — right up until the moment one of them lands on a hard floor from four feet, or something else lands on it.

Which means the entire prep is furniture, straps and where you put things. There is no setting to change and nothing to buy a subscription for. There is one genuine software exception, the offsite copy, and we will come to it, because it is the part that most people already believe they have handled and mostly have not.

The twenty minutes that does most of the work

The framework everybody in California teaches is "Secure Your Space," and the most concrete published version of it we have found is UC San Diego's workplace hazard guidance, which happens to translate almost line for line to a home office. It is worth reading its specifics rather than the general advice, because the specifics are what tell you which objects in your own room are the problem.

On the equipment itself, it says to secure "desktop computers, monitors, and printers" using "commercially available products including velcro fasteners and non-slip pads." That is the whole intervention for the things on your desk. A strip of hook-and-loop tape under a printer or a monitor base is a few dollars and stops the sliding that turns into a fall.

The line that matters most for the way modern home networks are actually laid out is this one: "attach raised shelf edges to prevent low-profile equipment from vibrating off the edge." Think about what is low-profile equipment in your house. It is the modem. It is the router. It is the network drive, the external backup disk, the little unmanaged switch, the streaming box. Almost all of it sits on a shelf, unattached, close to an edge, because that is where the cables reach. A lip along the front of that shelf — a strip of moulding, a length of dowel, anything that stands up an inch or two — is the single highest-return thing in this article.

For the furniture underneath, the rule it gives is a usefully mechanical one: anything over 42 inches tall whose height is more than about three times its base width should be bolted to the wall studs, or else moved so it cannot overturn onto people or block a doorway if it goes. That is most bookcases, most tall cabinets, and — very often — the narrow shelving unit somebody put in a corner precisely because the router had to go somewhere.

The rest is arrangement rather than hardware. Heavy things go on low shelves, because a heavy thing that falls three feet does far less than one that falls seven. Larger, heavier objects on open shelves can be restrained with eyehooks and a loop of cord, fishing line or thin wire. Cabinet doors get "positive latching devices" — slide bolts, safety hasps, the child-proof latches that close by themselves — so the doors do not swing open and empty the contents onto the floor.

None of that is exotic and none of it takes long. The reason it does not get done is not cost. It is that the room looks fine, and it will keep looking fine until the one afternoon it does not.

The desktop tower, the monitor, and the thing on the shelf above the desk

A few specifics for the equipment we actually get called about.

A desktop tower on the floor is in better shape than one on the desk, for the obvious reason that it has nowhere to fall. Its risk is different: it can walk, and if it walks it either tips or it puts strain on everything plugged into the back of it. The connectors that suffer are the ones with no flex — a short, taut power lead, a graphics card with a heavy cable hanging off it, an Ethernet run stapled to the skirting. Leave slack in the cables at the back of any machine that sits on the floor, and give the tower something to sit against that is not a route it can slide along.

A tower on a desk, or on the shelf under a desk with an open front, is the version worth moving. So is anything stored on a shelf directly above where a person sits, which in a lot of home offices is a row of heavy things over the chair.

Monitors are lighter than people assume and slide easily, so the non-slip pad under the base is genuinely worth the two minutes. A monitor on an arm clamped to the back edge of a desk is a different proposition — it is a weight on the end of a lever, and the clamp is doing all the work. It is worth checking that the clamp is actually tight, because they loosen over years of adjustment and nobody ever checks them.

And the printer, which is heavier than everything else on the desk and is almost always on a rolling cart or a side table with nothing holding it. Locked castors or a non-slip mat; it is not going to be destroyed by a fall, but it can take a lot of other things with it on the way down.

The two objects worth the most and secured the least

If you do nothing else in this article, do this bit.

The network drive — the NAS, if you have one — is the single worst-placed object in a typical home. It is small, it is dense, a four-bay tower model has a high centre of gravity for its footprint, and it lives on a shelf or on top of a cabinet rather than on a desk, because it makes noise and nobody wants it near them. It is also, in many households, the only place a decade of photographs exists at full resolution. It is precisely the object the raised-shelf-edge advice was written for, and it is the one nobody thinks to apply it to.

The second is the external backup drive on the desk, connected by one USB cable and nothing else. When the desk moves, the drive goes over the edge and either lands on the floor or hangs by its own cable, which is worse for the port. It weighs almost nothing, and the strip of velcro that holds it down costs less than the coffee you would buy on the way to fetch it.

There is one hardware point worth making here, because it is one of the few places where a purchasing decision genuinely changes the outcome. A spinning platter drive that is dropped while it is running is the bad case — the heads are flying microns above a spinning disc, and a sharp impact is the mechanism that turns a recoverable drive into a clean-room job. A solid-state drive has no moving parts and does not care. If you have been putting off replacing an old external disk, resilience is a better argument for the SSD than speed ever was, and this is the room where it applies.

The word "offsite" is doing far more work than people realise

Now the software half, and the reason this article exists at all rather than being a link to the backup one.

We have written the 3-2-1 rule out properly elsewhere: three copies of anything important, on two kinds of media, with one of them offsite. It is the right rule and we stand behind it. But the offsite copy is the part that quietly fails in earthquake country, and it fails for a reason that never comes up with any other hazard.

Ask people where their offsite copy is and the common answers are: a drive at my sister's house, a drive in my desk at the office, a second machine at the shop. Every one of those is a genuinely good offsite copy — for a house fire, a burst pipe, a burglary, or a ransomware infection that only reaches machines on one network. All of those hazards are one building wide. Put the copy in a different building and you have beaten them.

An earthquake is not one building wide. It is a region wide. Your house in Northridge and your sister's house eight miles away in Burbank are not two independent locations for this hazard; they are two addresses inside the same event, on the same rough patch of ground, quite possibly served by the same utility and reached by the same freeway.

The disaster-recovery industry, which has thought about this for decades because it has to, puts a number on it. The figure that recurs across continuity guidance is a minimum of about 100 miles between a primary site and its backup site, on a different power grid and different network infrastructure — with 300 miles or more recommended for organisations in the highest-risk categories, and California earthquake zones are explicitly named among them. The published caution is blunt: geographic distance alone is not the point, because sites can be far apart and still share a grid or a network path, and separation that looks diverse on a map has repeatedly turned out not to be.

That guidance is written about data centres and it is obviously not asking you to drive a hard drive to Nevada. But the geography argument underneath it is exactly the same at household scale, and it collapses to one sentence: a cloud backup is offsite by construction, and a drive you can reach in a twenty-minute drive is not, for this particular hazard.

The honest version of the advice, which the vendor blogs will not give you because it is not what they are selling, is that you should keep the local drive. It is the fast restore. It is the copy you will actually use, because dead drives, deleted folders and failed upgrades are enormously more common than major earthquakes and the local copy handles all of them in an afternoon. Keep it, use it, put a shelf lip in front of it. Just stop counting it as your offsite copy, because for this one it is not, and if it is the only other copy you have then you do not have the third leg of the rule at all.

The small-business version, which has a problem the house version does not

For a shop, an office or a practice, everything above applies, plus one thing that changes the whole calculation.

The equipment list is a little different — a register and a card terminal, a back-office machine that quietly runs the whole operation, a small server or a NAS on a shelf in a stockroom, and a shelving unit full of display stock next to all of it. All of that gets the same treatment: lips on the shelves, the tall units bolted to studs, the heavy stock down low, latches on the cabinets.

The difference is access. After a significant quake, a building can be inspected and tagged, and if it is tagged you may not be allowed back inside for days or weeks. That is the scenario the household version does not have to think about, and it turns a data problem into an availability problem. Your backup can be perfectly intact, sitting on a shelf, undamaged — and completely useless to you, because it is inside a building you are not permitted to enter.

So the test for a business is not "do I have a backup." It is narrower and much more uncomfortable: could you, today, from a laptop somewhere else, without going into your building, reach the things that keep the business alive? The customer list. The books. Payroll. The supplier contacts. The passwords. If the honest answer for any of those is "it is on the machine in the back," that is the thing to fix in the next six weeks, and it is fixed with a cloud copy rather than with a strap.

The other item worth five minutes is the paper: whatever you would need to prove what you had. Take photographs of the equipment and the serial numbers, and keep them somewhere that is not the building — an insurer will want them, and afterwards is a bad time to discover the only inventory was a spreadsheet on the office desktop.

Turn the alerts on — and know what they can and cannot do

This is a phone setting rather than a computer job, it takes about ninety seconds, and in the households we visit it is off about as often as it is on. It belongs here because it is the only part of earthquake preparedness that is genuinely technological, and because there is a lot of confusion about what it actually is.

California's earthquake early warning programme lists three ways to receive alerts. The MyShake app, which is free and runs on iOS, Android, Chromebook and macOS. Android's built-in earthquake alerts, which are included automatically on millions of Android phones in California through a partnership with Google and need no app at all — worth knowing, because a lot of people install the app on a phone that was already covered. And Wireless Emergency Alerts, the nationwide system that sends short messages to any WEA-capable handset.

The thresholds are published and worth knowing so you are not surprised by the silence. Alerts go out for earthquakes of magnitude 4.5 or greater where the expected shaking reaches Modified Mercalli Intensity III — "weak" — or above. The MyShake app will show you information about smaller events, from about magnitude 2.5, but that is display, not an alert.

What you get is described by the state as "seconds to tens of seconds" of warning, and the reason for that range is the honest limitation of the whole system: the alert travels at the speed of a data network and the shaking travels at the speed of rock, so the further you are from the origin the more warning you get. The corollary is stated plainly on the state's own page — people very close to the epicentre may receive the alert while the shaking is already happening, or after it has passed. It is a genuinely useful few seconds for most of the region and it is not a promise.

One thing it is not: a trigger for your house. The alert is for people — it exists so that a human being can drop, cover and hold on. Nothing in your home is going to act on it, and it is not a substitute for any of the physical work above. Get it on every phone in the household, then go back to the shelves.

Afterwards: the first hour, and the one thing not to do

The instinct once the shaking stops and everyone is accounted for is to start plugging things back in to see whether they still work. Give it the ten minutes it takes to look first, because two of the mistakes available here are expensive and both of them are made in the first hour.

Look before you power. Anything that fell gets inspected before it is switched on — a cracked case, a bent or partly pushed-in connector, a power supply that now rattles, a laptop with a swollen or dented battery. A bent connector energised is a short; the point of the inspection is to catch it before you supply the electricity.

And the one genuine "do not" of the whole article: a hard drive that fell while it was running, and that holds something you actually need, should be left alone. Do not keep powering it up to see if it comes back this time. That is the same rule we give for a drive that has been wet, and for the same underlying reason — every additional spin-up on a drive with a damaged head assembly is another chance to convert a recoverable disc into an unrecoverable one, and the recovery you might have paid a few hundred dollars for becomes one you cannot buy at any price. If it is clicking, grinding or not being detected, stop and let someone look at it. We have written the warning signs up separately.

While you are walking around, look up as well as down. A pipe or a sprinkler head that has been shaken loose above a shelf of equipment is a slow-motion version of a problem we have written about at length, and it does its damage over hours rather than seconds, so it is the one worth catching early.

Then there is power and internet, which will both come back untidily. We have covered that ground already and will not repeat it here: what to check when the power comes back and the computer does not, and the half of your home internet that fails in an outage even when the provider's network is fine. Both apply directly and both are linked below.

The weekend before October 15

If you want the whole thing as a list, here it is. It is about twenty minutes of work and one decision that costs money.

One: put a lip, a strip of velcro, or both, under and in front of the low-profile gear — the modem, the router, the network drive, the external backup disk, the switch. Two: bolt the tall bookcase or shelving unit that this equipment lives on, or next to, to the wall studs. Three: move the heavy things off the high shelf above wherever somebody sits, and put a non-slip pad under the monitor and the printer. Four: turn on earthquake alerts on every phone in the house, and check the Android ones are not already on before installing anything.

Five, and this is the one that costs money: work out whether you have a copy of your data that is genuinely out of the region, and if you do not, start one now, because the first upload of a large photo library takes days rather than minutes and there is no point starting it on October 14. Then restore one file from it, so you know it works. A backup nobody has ever restored from is not a backup yet.

Six, for a business: write down what you would need to reach if you could not get into the building, and check that you can actually reach all of it from somewhere else.

The drill itself takes a minute and is worth doing with whoever is in the house or the shop, for the same reason fire drills are worth doing — the useful part is not the information, it is having done it once before it mattered.

How we can help

We set up backups for homes and small businesses across Southern California every week, and this is a normal service call rather than anything specialised: we look at where the equipment actually sits, sort out the copy that is genuinely offsite, and — the step people skip — restore a file from it in front of you so you know it works. If you would rather not hand your data to a subscription, we also build backup systems that run on storage you own with redundancy across drives.

For a shop or an office we will do the same thing plus the access question: making sure the books, the customer list and the credentials are reachable from a laptop somewhere else, so a building you cannot get into is an inconvenience rather than the end of the business.

And if something has already fallen — a drive that is clicking, a machine that will not start, a network box that took the shelf with it — bring it in or call before you keep trying it. The first hour is where the recoverable cases get lost.