Free Space Is the Gate Before Every Update: Two Capacity Fields and the Arithmetic on Batching

Published 2026-11-08 · LuckyMDM Blog

The one-line version: an OS update that will not install, an enterprise app that will not deploy, and a re-provision that fails after erasure all look unrelated in a console, but on the device they hit the same wall — free space. Only two numbers describe that wall: total capacity and available capacity, both in gigabytes, not bytes and not a percentage. The part that trips people up is that available is not the same as reclaimable, so the figure works as a gate and fails as a cleaning target.

Two capacity numbers, and only two

`DeviceCapacity` is what the box shipped with

`DeviceCapacity` returns total capacity in gigabytes: 64 for a 64 GB unit, 256 for a 256 GB unit. It barely moves during a device's life. That makes it useful for grouping — scheduling by storage tier — and useless for deciding whether something can be installed right now.

`AvailableDeviceCapacity` is the remainder at the moment of the query

`AvailableDeviceCapacity` returns available capacity, also in gigabytes, supported from iOS and iPadOS 4. It means the remainder at the instant the query ran, not the theoretical space a cleanup could reach. Those are very different claims: the first is what the device has free, the second only becomes true after someone deletes something.

The difference is occupancy, not reclaimable space

QuantitySource fieldUnitGood forNot good for
Total capacity`DeviceCapacity`GBGrouping, batching by tierDeciding if an install fits
Available capacity`AvailableDeviceCapacity`GBSpace gate before an updateEstimating what a cleanup will free
OccupancyDifference of the twoGBUtilisation reportingTreating as reclaimable

The third row is where most mistakes live. Read "52 GB occupied" as "52 GB recoverable" and every cleanup you plan will come in short.

LuckyMDM (Sichuan Starlight Network LLC) stores `DeviceCapacity` and `AvailableDeviceCapacity` as two separate ledger fields and stamps every query result with its collection time — because the premise for any install decision is a remainder from a known moment, not a number of unknown age.

Why a large part of occupancy cannot move

The mechanism here has little to do with file sizes. This space is held by structure, not by files. Three places account for it.

The system volume takes a fixed share

A 64 GB iPhone typically shows around 45 GB available after a full erase and reset. The missing share is the system volume and bundled apps. It does not shrink when a user deletes photos, and it does not disappear after erasure — so "erase it and you get 64 GB back" is not a reachable target.

APFS snapshots and system data are not file sizes

iOS runs on APFS, where snapshots, clones and what Settings presents as "System Data" do not map to the size of any single app. The storage screen may show one lumped number while the actual footprint is spread across dozens of system containers. This is why deleting a batch of photos sometimes moves the available figure by less than expected.

Deleting an app frees the app, not its footprint

Removing an app releases its bundle and its user data. Its caches, files it exported into the Files app, and media it wrote into the photo library stay behind. An estimate of "delete three big apps and free 20 GB" therefore tends to land at roughly half.

An update needs two copies of the package

Download, then unpack

An over-the-air update happens in two stages: the package downloads to local storage, then it is unpacked and installed. Both happen on the same volume, so peak usage runs well above the package size alone.

The working figure: about three times the package

For a 4 GB package, peak usage sits around 8 GB, and leaving headroom for logs and temporary files during installation brings a safe threshold to 12 GB. That is not a published Apple constant — it is an engineering figure derived from two to three times the package size. The underlying reason is straightforward: the download and the unpacked working set coexist on one volume, so the peak must exceed the package. A multiplier also survives generation changes, which a fixed gigabyte number does not.

Three shapes of failure

The first is failure at download, usually surfaced as "unable to check for updates", which reads like a network problem. The second is failure during installation: the unit reboots, the version does not change, and a chunk of space is gone. The third is a stall mid-install with the device looping and never reaching the home screen — the expensive one, because recovery means a restore.

All three land at different points on the same chain, yet the console shows three unrelated symptoms: a command timeout, an unchanged build number, and a device that stops checking in. Read only from the console, it is hard to group them as one problem class.

The arithmetic on 500 units

18 percent short is 90 units

Take a 500-unit fleet, mostly 64 GB models, deployed for six months or more, with 18 percent of units sitting below 12 GB available. That is 90 units that will fail the update.

90 units at 12 minutes is 18 hours

The handling loop per unit: check capacity, clear caches, remove non-essential apps, get above 12 GB, re-push, wait for a receipt. At 12 minutes each, 90 units is 18 hours of labour. Compressed into an update window, that stretches the window to three working days.

ApproachCompletionLabourRework
Push to all 500 at onceStarts at night, 90 units fail all night18 hours concentrated the next dayUpdate package downloads again, using more space
Batch by capacity410 units finish that night18 hours spread across two daysNone

Total labour is identical. The difference is the window: a single push dumps every failure into the next working day, which usually already has deliveries on it. Batching splits the same 18 hours into two nine-hour days and keeps the delivery day intact.

The point is not saving labour. It is converting an uncontrolled "everything fails overnight" into a planned "clean up first".

LuckyMDM's fleet ledger keeps the space gate ahead of scheduling rather than behind failure: group first, read the remainder second, set batches last — reverse the order and the cost moves from planning into rework.

Three checks you can run yourself

Check one: look at the real distribution on a device

Path: Settings > General > iPhone Storage. This screen ranks apps by size and shows a "System Data" row. Watch that row specifically — it is the part that will not clear by deleting apps. If System Data alone holds a large share, the return on cleanup work is limited.

Check two: query two fields, not one

Ask `DeviceInformation` for both `DeviceCapacity` and `AvailableDeviceCapacity`. A remainder on its own loses the tier: 10 GB left is close to critical on a 64 GB unit and unremarkable on a 256 GB unit.

Check three: run an erase comparison

Pick 3 units and record available capacity. Erase, skip the data restore, and record it again. The gap is what cleanup actions can actually free on that model — one run gives you a local figure you can plan against instead of a guess.

Boundaries and where this does not apply

Boundary one: `AvailableDeviceCapacity` is not in the tvOS query set. It is available on iOS, iPadOS, macOS, visionOS and watchOS, but not on Apple TV. A script that queries every platform will report a missing field on tvOS — that is a support boundary, not a device fault.

Boundary two: enterprise app footprint needs different fields. To measure what an MDM-installed app occupies, use `BundleSize` and `DynamicSize` from `InstalledApplicationList`, both in bytes. Estimating from an App Store listing misses runtime data and typically lands 30 percent to 50 percent low.

Boundary three: Shared iPad reports a different unit. On multi-user devices each user has a quota, returned as `Storage quota size` in megabytes, not gigabytes. Judging a Shared iPad with a single-user capacity figure produces the wrong answer.

Boundary four: available capacity is a snapshot, not a trend. A single query describes one moment. Building a trend requires sampling at a fixed interval with timestamps; two readings a month apart may straddle an OS update and mislead.

Boundary five: the erase preflight check is macOS-only and version-gated. On macOS, whether Erase All Content and Settings can run is reported by `EACSPreflight`, available from macOS 13.3. On earlier systems you cannot pre-judge from the field.

Common misconceptions

Misconception one: total capacity minus installed apps equals free space

That arithmetic ignores photos, system data, snapshots, caches and offline media. A unit with 20 apps totalling 15 GB can show far less than "64 minus 15" available. The error comes from treating occupancy as the sum of visible apps.

Misconception two: erasing frees everything

Erase resets the data volume; the system volume remains. Around 45 GB available on a 64 GB model after a full erase is normal, not evidence of an incomplete wipe. Expect 64 GB and you will erase the same unit repeatedly, chasing a number that will not move.

Misconception three: a percentage is enough

Ten percent of 64 GB is about 6 GB; ten percent of 256 GB is about 25 GB — a fourfold difference. A percentage-based alert therefore never fires on large units and fires constantly on small ones. Thresholds in gigabytes are comparable; thresholds in percent are not.

Misconception four: deleting apps is the most efficient cleanup

On subscription fleets the units usually carry few user apps. The bulk sits in photos, video and system data, so clearing the photo library and caches returns more than removing apps.

FAQ

Why is available capacity so much lower than the advertised size?

Because the system volume, bundled apps and recovery partition occupy space outside the user-visible share. Around 45 GB available on a 64 GB model after an erase is the normal range, independent of usage habits.

An update will not install. How do I tell space from policy?

Two signals. First, whether available capacity is below twice the package size. Second, whether the failure happened at download or at install — download failures are usually network or storage, install failures are almost always space. Policy problems usually look like the command never being delivered, not being delivered and failing.

How much free space is enough?

Set the gate at three times the package size: 12 GB for a 4 GB package. Below twice the package size, installation fails intermittently, and a failed install leaves the device in a state that is harder to recover than a failed download.

Should capacity alerts be set by percentage or by gigabytes?

By gigabytes, per storage tier. Something like 8 GB for 64 GB units and 20 GB for 256 GB units keeps the alert meaningful across the fleet.

After an erase the space still looks short. Is the device faulty?

Compare against the model baseline first. If available capacity is clearly below the average for that model, storage may need attention. If it simply never reaches the advertised total, that is ordinary system occupancy, not a fault.

What to check before you schedule anything

All Articles