Struct LifecycleManager 

pub struct LifecycleManager { /* private fields */ }

Per-bucket lifecycle configuration manager.

All read / write operations go through RwLock for thread safety; clones are cheap (Arc<LifecycleManager> is the expected handle shape). actions_total is a parallel RwLock<HashMap<...>> of (bucket, action_label) -> count so the future background scanner can stamp successful actions and operators can GET /metrics to see the running totals (the metric is also surfaced via metrics::counter! — see crate::metrics::record_lifecycle_action).

Implementations

impl LifecycleManager

pub fn new() -> Self

Empty manager — no bucket has rules.

pub fn put(&self, bucket: &str, config: LifecycleConfig)

Replace (or create) the lifecycle configuration for bucket. Drops any previously-attached rules in one shot — matches AWS S3 PutBucketLifecycleConfiguration (full replace, no merge).

pub fn get(&self, bucket: &str) -> Option<LifecycleConfig>

Return a clone of the bucket’s configuration, if any.

pub fn delete(&self, bucket: &str)

Drop the bucket’s lifecycle configuration (idempotent — missing bucket is OK).

pub fn to_json(&self) -> Result<String, Error>

JSON snapshot for restart-recoverable state. Pair with Self::from_json.

pub fn from_json(s: &str) -> Result<Self, Error>

Restore from a JSON snapshot produced by Self::to_json. Action counters are intentionally not snapshotted — they’re transient observability data and should reset across process restarts so rate(s4_lifecycle_actions_total[1h]) doesn’t double-count.

pub fn evaluate( &self, bucket: &str, key: &str, object_age: Duration, object_size: u64, object_tags: &[(String, String)], ) -> Option<LifecycleAction>

Evaluate which rule (if any) applies to a single current-version object right now. Walks the bucket’s rules in declaration order; returns the first matching action. Returns None when no rule matches (or when the matching rule is Disabled, or when the bucket has no lifecycle configuration).

Within a single rule the precedence is:

1. Pick the deepest transition whose days threshold is currently met (= largest days ≤ object age).
2. Conflict with expiration: if expiration_days <= transition_days for the chosen transition, expiration wins (the rule wants the object gone before it would have been transitioned). Otherwise transition wins (e.g. transition at 30d, expiration at 365d, age 60d → transition fires now, expiration is future).
3. expiration_date matches when now >= expiration_date and no transition is currently applicable.

object_age is “now - created_at” supplied by the caller — keeping the evaluator pure of the wall clock makes deterministic testing trivial.

pub fn evaluate_with_flags( &self, bucket: &str, key: &str, object_age: Duration, object_size: u64, object_tags: &[(String, String)], flags: EvaluateFlags, ) -> Option<LifecycleAction>

Full-form evaluator with flags for noncurrent-version handling. Use this when the scanner is walking a versioning-enabled bucket; pass is_noncurrent = true for entries that are not the latest non-delete-marker version.

pub fn evaluate_in_flight_multipart( &self, bucket: &str, upload: &MultipartUploadCandidate, now: DateTime<Utc>, ) -> Option<LifecycleAction>

v0.8.3 #69 (audit M-2): evaluate one in-flight multipart upload against the bucket’s rules. Returns LifecycleAction::AbortMultipartUpload when at least one Enabled rule (a) accepts the upload’s key via its filter and (b) carries an abort_incomplete_multipart_upload_days threshold whose age (now - initiated) is currently met. Returns None otherwise (no matching rule, no abort-multipart-upload-days set, or the upload is too young).

Filter matching reuses LifecycleFilter::matches with object_size = 0 — in-flight uploads have no assembled size yet (the parts are stored independently in the backend), so any rule whose filter sets object_size_greater_than / object_size_less_than is treated as if the upload were 0 bytes. AWS S3 itself does not gate AbortIncompleteMultipartUpload on size; this matches the AWS semantic (size predicates simply do not apply to the abort path) for the typical filter shape (no size predicate). Operators wanting size-gated abort can carry the upload’s declared part length on the MultipartUploadCandidate in a follow-up issue — the API extension is additive.

pub fn record_action(&self, bucket: &str, action: &LifecycleAction)

Stamp the per-bucket action counter and bump the matching Prometheus counter. Called by the future scanner after a successful delete / metadata rewrite.

pub fn actions_snapshot(&self) -> HashMap<(String, String), u64>

Read-only snapshot of the per-(bucket, action) counter map. Useful for tests + introspection (/admin/lifecycle/stats style endpoints in the future).

pub fn buckets(&self) -> Vec<String>

All buckets with a lifecycle configuration attached. Sorted for stable scanner ordering.

Trait Implementations

impl Debug for LifecycleManager

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for LifecycleManager

fn default() -> LifecycleManager

Returns the “default value” for a type.