Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

Turned on BitLocker and the next backup was a full copy: why, and how to plan the one-off growth

Technical article
Backup & Recovery

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

The next backup is only a full copy if it reads the disk in blocks. Turning on BitLocker makes Windows read and rewrite every sector it encrypts, so a backup that asks the hypervisor "which blocks changed?" is told that all of them did. A backup that reads files through Windows sees the decrypted data and notices nothing.

The jump itself happens once; whether the daily growth then returns to its old size depends on how your backup stores encrypted blocks. The practical job is to confirm which kind of backup you have, work out how much extra space the one-off needs, and decide when it lands.

What the symptom means

BitLocker encrypts a volume under Windows and keeps it readable to Windows. Microsoft's FAQ describes the two sides: "encrypted sectors in the BitLocker-protected drive are decrypted only as they're requested from system read operations", and "blocks that are written to the drive are encrypted before the system writes them to the physical disk."

Those two statements decide the backup outcome:

  • A file-level backup, or an agent inside the guest that reads files, goes through Windows. It receives decrypted data. Enabling BitLocker does not change which files changed.
  • A block-level backup reads the disk below Windows. VMware's virtual disk documentation describes changed block tracking as virtual machines that "keep track of disk sectors that have changed", with a query that "returns the disk sectors that changed" since a saved change ID. Hyper-V's WMI backup uses resilient change tracking, which the Hyper-V backup documentation lists next to a Win32 QueryChangesVirtualDisk call. Encryption rewrites the sectors it converts, and those sectors now hold different bytes, so they are changed sectors.

There is one more point about size. Microsoft's manage-bde -on page states that with the -usedspaceonly option "the sections of the volume containing used space will be encrypted but the free space will not", and that "if this option is not specified, all used space and free space on the volume will be encrypted." So a full-volume encryption touches the entire volume, and used-space-only touches the used part.

Because no unencrypted data is ever stored on a BitLocker-protected drive (BitLocker FAQ), a block-level backup reads encrypted blocks. Encrypted data generally compresses and deduplicates poorly, so each later daily change can also be stored larger for as long as the volume stays encrypted. Microsoft's pages give no figure, so measure the first fortnight; the linked article on compressed and encrypted files covers why.

Likely causes

Cause What you see How to confirm
Block-level backup and the volume was fully encrypted One backup stores about the size of the whole volume, free space included manage-bde -status shows a conversion that finished around the date of the jump; the job reads the virtual disk, not files
Block-level backup and used-space-only encryption One backup stores about the used size of the volume Same check; compare the one-off to used space, not to disk size
Encryption method changed (decrypt, then encrypt again) Two large backups, one for each pass Microsoft says an already-encrypted device "must first be decrypted before the new encryption method" applies, so two rewrites happen
Shadow copies made before BitLocker was enabled were deleted Earlier shadow-copy restore points are gone The FAQ says shadow copies made before enabling are deleted on software-encrypted drives; hardware-encrypted drives keep them
Something else reset change tracking at the same time Full read, but the volume is not encrypted or the dates do not line up See the linked article on change tracking resets

Prove the cause before changing anything

These checks only read. Start with the one that separates the two backup types.

  1. Where does your backup read? Find out whether the job runs an agent inside Windows that reads files or snapshots, or reads the virtual disk from the hypervisor. If it reads files, BitLocker is not the cause of a full copy and the last row of the table is where to look. Your backup tool's job settings or the provider's documentation say which.
  2. What state is the volume in?
manage-bde -status C:

Microsoft documents that manage-bde -status reports size, BitLocker version, conversion status, percentage encrypted, encryption method, protection status, lock status, identification field and key protectors. The page shows no sample output, so compare the lines it prints against that list. A conversion that is still running means the backup will keep seeing newly rewritten sectors night after night until it reaches the end. Omit the drive letter to list every drive.

  1. Does the date match? Compare the date the conversion started or finished with the first oversized backup, using the backup tool's own report of data transferred or blocks changed. A jump on the same night points to this cause. A jump on another night points elsewhere.
  2. Size the jump. If the backup report shows the amount changed, compare it with the volume's used space and with its full size. Close to used space suggests used-space-only encryption; close to total size suggests full-volume encryption.

Example (illustrative): fs01 is a virtual machine with a 500 GB volume holding 300 GB of data that changes 1% (3 GB) a day. The backup keeps 30 points and stores changed blocks once. Before BitLocker, the store holds 300 GB for the first point plus 29 x 3 = 87 GB, a total of 387 GB. If the encryption is full-volume and the backup stores every changed block at full size, the night after conversion adds up to 500 GB, and the store reaches about 387 + 500 = 887 GB. With used-space-only encryption it adds up to 300 GB, and the store reaches about 387 + 300 = 687 GB. Both are upper bounds; the old unencrypted points stay until they age out of retention, so the store holds both for a while.

Fix

There is nothing to repair. The aim is to land the one-off deliberately.

Plan the one-off

  • Make sure the backup target has the extra space from the example above before the conversion starts. Work from the used-size or volume-size figure that matches the encryption type.
  • Convert one volume or server at a time so one night does not carry several one-offs.
  • Do it before a volume is first backed up where you can. A server whose first backup is already encrypted pays nothing extra.

Choose the encryption type on purpose

Microsoft's planning guide says full-drive encryption "is useful for drives that have been repurposed, and might contain data remnants from their previous use", and used-space-only is faster and encrypts data as it is saved. Its caution is the other side of the choice: on a volume that already held confidential data, used-space-only leaves earlier data in free space recoverable "until they're overwritten". For backup size, used-space-only is smaller; for security on an old volume, full may be required. Decide with the data owner.

Spread the work if one night is too heavy

manage-bde -pause C: pauses encryption or decryption, and manage-bde -resume C: continues it (Microsoft's FAQ adds that encryption also resumes after a restart where it stopped). Pausing does not reduce the total, it moves it.

Risk: A paused conversion leaves a volume partly encrypted until you resume it, and each resume rewrites more sectors that the next backup will see. If the pause is forgotten, the one-off becomes a slow trickle that hides inside normal change. Back out by running manage-bde -resume C:. Note the date you paused and the date you resumed.

Do not undo it to "fix" the backup

Risk: Decrypting a volume is a second full rewrite, so it causes a second large backup, and turning BitLocker on again causes a third. The FAQ states that decrypt "completely removes BitLocker protection and fully decrypts the drive". If you are tempted to decrypt, keep the recovery information first; the linked recovery-keys article covers that. The only way back from a decryption is to encrypt again, which is another full rewrite.

Confirm the fix

  1. manage-bde -status C: shows the conversion has finished (percentage encrypted at the end of the range).
  2. The next two backups after conversion report a changed-data volume far below the volume size. Compare it with the old daily figure; it may sit above it for as long as the volume stays encrypted.
  3. The target shows the extra space in use, and the old points start to age out as retention passes.

What to watch afterwards

  • A later change of encryption method forces a decrypt and re-encrypt, so schedule it like a new conversion.
  • BitLocker enablement on new servers by policy can cause a one-off per server on the same night. Stagger it.
  • The extra space stays in use until the points taken before encryption age out of retention, so check the target's free space for that whole period, not only on the night of the jump.

Risks and limits

  • These commands have not been run or lab-tested. They come from Microsoft's manage-bde and FAQ pages.
  • The manage-bde -status page lists the fields it reports but shows no output, so its exact layout is not described here.
  • Microsoft's and VMware's pages do not say in so many words that a changed-block query returns the sectors rewritten by BitLocker. That follows from the change-tracking descriptions combined with the BitLocker statements above, and it is the part to confirm with your backup provider.
  • Whether a particular backup product stores unallocated or zero blocks at full size is not documented in these sources. The example uses upper bounds for that reason.
  • Behaviour on hardware-encrypted drives differs; Microsoft only notes that shadow copies are kept there.
  • Figures in the example are made up to show the arithmetic.

Sources