TQ: Support for ZFS Key Rotation #9737
Draft
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
When Trust Quorum commits a new epoch, all encrypted U.2 datasets need
their encryption keys rotated. This change implements that flow:
- trust-quorum: Add watch channel to broadcast committed epoch changes
from NodeTask to subscribers
- sled-agent: Wire committed_epoch_rx to the config reconciler
- config-reconciler:
- Add KeyRotationError, RekeyRequest types for the rekey API
- Add rekey_datasets() batch operation on DatasetTaskHandle
- Add datasets_rekey() to DatasetTask for serialized key rotation
- Add rekey_for_epoch() to OmicronDatasets to coordinate rekeying
all managed disks when an epoch is committed
- Handle epoch change notifications in the reconciler run loop
- Add managed_disks() iterator to ExternalDisks
- illumos-utils:
- Add Zfs::change_key() using zfs-atomic-change-key crate
- Add ChangeKeyError type
- Add epoch field to DatasetProperties and include oxide:epoch in
ZFS property queries
- key-manager: Add Debug derives to key types
The rekey operation is idempotent: datasets already at the target epoch
are skipped. On startup, we process the initial epoch to catch any
missed rekeys from crashes.
Fixes #9587
plaidfinch
commented
Jan 28, 2026
|
|
||
| /// Derived Disk Encryption key | ||
| #[derive(Default)] | ||
| #[derive(Debug, Default)] |
Contributor
Author
There was a problem hiding this comment.
We use this in a struct that implements Debug elsewhere. SecretBox has a [REDACTED] impl, so this is safe.
Comment on lines
+297
to
+302
| pub(super) fn managed_disks(&self) -> impl Iterator<Item = &Disk> { | ||
| self.disks.iter().filter_map(|disk_state| match &disk_state.state { | ||
| DiskState::Managed(disk) => Some(disk), | ||
| DiskState::FailedToManage(_) => None, | ||
| }) | ||
| } |
Contributor
Author
There was a problem hiding this comment.
Should we return an error if we hit the FailedToManage state anywhere, or — as is done here — silently omit any such disk?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When Trust Quorum commits a new epoch, all encrypted U.2 datasets need their encryption keys rotated. This change implements that flow:
trust-quorum: Add watch channel to broadcast committed epoch changes fromNodeTaskto subscriberssled-agent: Wirecommitted_epoch_rxto the config reconcilerconfig-reconciler:KeyRotationError,RekeyRequesttypes for the rekey APIrekey_datasetsbatch operation onDatasetTaskHandledatasets_rekeytoDatasetTaskin the ZFS operation serializer task for key rotationrekey_for_epochtoOmicronDatasetsto coordinate rekeying all managed disks when an epoch is committedmanaged_disksiterator toExternalDisksillumos-utils:Zfs::change_keyusingzfs-atomic-change-keycrate (temporarily) to rotate keys atomically with the change of theoxide:epochpropertyChangeKeyErrortypeepochfield toDatasetPropertiesand includeoxide:epochin ZFS property querieskey-manager: Add Debug derives to key typesThe rekey operation is idempotent: datasets already at the target epoch are skipped. On startup, we process the initial epoch to catch any missed rekeys from crashes.
Fixes #9587