Compare commits

..

5 Commits

8 changed files with 576 additions and 317 deletions
-1
View File
@@ -41,7 +41,6 @@ script-tests: ## Run shell script tests
$(RUSTFS_PYTHON_BIN) ./scripts/check_security_coverage.py --self-test
$(RUSTFS_PYTHON_BIN) ./scripts/check_scheduled_validation_freshness.py --self-test
$(RUSTFS_PYTHON_BIN) ./scripts/test_security_workflow.py
$(RUSTFS_PYTHON_BIN) ./scripts/test_nightly_candidate.py
$(RUSTFS_PYTHON_BIN) ./scripts/s3-tests/test_report_compat.py
bash -n ./scripts/validate_object_data_cache_cold_stampede.sh
$(RUSTFS_PYTHON_BIN) ./scripts/check_object_data_cache_follower_samples.py --self-test
+5 -5
View File
@@ -10,16 +10,16 @@ Use N/A when there is no related issue.
## Summary of Changes
<!--
Describe the concrete problem and resulting behavior. For a behavior change, name the input or state that triggers it and the expected outcome. Explain any new dependency or abstraction that the change needs.
Briefly explain what changed and why reviewers should accept it.
Focus on behavior, compatibility, and review-relevant context.
-->
## Verification
<!--
Give 13 concrete pieces of evidence for the changed behavior: the test or command, its observed result, and the regression it catches. For a bug fix, record a failing-before/passing-after check or explain why it was unavailable.
List the commands or checks you ran, for example:
- `make pre-commit`
Identify the tested commit and any local changes. When testing a prebuilt binary or external service, include its source/version and artifact identity; a successful run against a different build is not evidence for this change.
List relevant checks not run and the remaining risk. Use the validation tier in AGENTS.md; do not run broader checks solely to fill this section. For documentation-only changes, list the applicable documentation checks. Use N/A only when verification is not applicable.
Use N/A only when verification is not applicable.
-->
## Impact
-1
View File
@@ -168,7 +168,6 @@ jobs:
python3 ./scripts/check_test_wiring.py --self-test
python3 ./scripts/check_scheduled_validation_freshness.py --self-test
python3 ./scripts/test_security_workflow.py
python3 ./scripts/test_nightly_candidate.py
python3 ./scripts/check_test_wiring.py
- name: Check no planning docs committed
+8 -51
View File
@@ -166,9 +166,8 @@ jobs:
# e.g. https://dl.rustfs.com/artifacts/rustfs/packages/nightly/... .
# Skipped when the R2 secrets are not configured (artifact-only mode).
- name: Upload DEB to Cloudflare R2
id: publish
if: env.R2_ACCESS_KEY_ID != ''
env:
DEB_FILE: ${{ steps.deb.outputs.deb_file }}
R2_ACCESS_KEY_ID: ${{ secrets.R2_ACCESS_KEY_ID }}
R2_SECRET_ACCESS_KEY: ${{ secrets.R2_SECRET_ACCESS_KEY }}
R2_ENDPOINT: ${{ secrets.R2_ENDPOINT }}
@@ -183,70 +182,28 @@ jobs:
exit 0
fi
if ! command -v aws >/dev/null 2>&1; then
sudo apt-get update && sudo apt-get install -y -qq awscli
fi
export AWS_ACCESS_KEY_ID="$R2_ACCESS_KEY_ID"
export AWS_SECRET_ACCESS_KEY="$R2_SECRET_ACCESS_KEY"
export AWS_DEFAULT_REGION="auto"
SOURCE_SHA="$(git rev-parse HEAD)"
if [[ "${SOURCE_SHA}" != "${GITHUB_SHA}" ]]; then
echo "Checkout SHA does not match the nightly build run" >&2
exit 1
fi
DEB_SHA256="$(sha256sum "${DEB_FILE}" | cut -d ' ' -f 1)"
CANDIDATE_KEY="artifacts/rustfs/packages/nightly/runs/${GITHUB_RUN_ID}/${GITHUB_RUN_ATTEMPT}/${DEB_SHA256}/rustfs.deb"
CANDIDATE_URL="https://dl.rustfs.com/${CANDIDATE_KEY}"
# Old AWS CLI models lack conditional PutObject support. Never fall
# back to an overwriting upload for a candidate.
AWS_CLI=aws
if ! "${AWS_CLI}" s3api put-object --generate-cli-skeleton input | jq -e 'has("IfNoneMatch")' >/dev/null; then
sudo apt-get update
sudo apt-get install -y -qq python3-venv
AWS_CLI_DIR="$(mktemp -d "${RUNNER_TEMP}/nightly-awscli.XXXXXX")"
trap 'rm -rf "${AWS_CLI_DIR}"' EXIT
python3 -m venv "${AWS_CLI_DIR}"
"${AWS_CLI_DIR}/bin/python" -m pip install --disable-pip-version-check 'awscli==1.44.79'
AWS_CLI="${AWS_CLI_DIR}/bin/aws"
fi
"${AWS_CLI}" s3api put-object --generate-cli-skeleton input | jq -e 'has("IfNoneMatch")' >/dev/null
"${AWS_CLI}" --version
"${AWS_CLI}" s3api put-object --bucket "${R2_BUCKET}" --key "${CANDIDATE_KEY}" \
--body "${DEB_FILE}" --if-none-match '*' --endpoint-url "${R2_ENDPOINT}"
PUBLISHED_SHA256="$(curl -fsSL --retry 3 --connect-timeout 15 --max-time 300 "${CANDIDATE_URL}" | sha256sum | cut -d ' ' -f 1)"
if [[ "${PUBLISHED_SHA256}" != "${DEB_SHA256}" ]]; then
echo "Published candidate checksum does not match the built package" >&2
exit 1
fi
DEB_FILE="${{ steps.deb.outputs.deb_file }}"
R2_PREFIX="s3://${R2_BUCKET}/artifacts/rustfs/packages/nightly/"
echo "📤 Uploading ${DEB_FILE} to ${R2_PREFIX}"
"${AWS_CLI}" s3 cp "${DEB_FILE}" "${R2_PREFIX}" --endpoint-url "$R2_ENDPOINT" --only-show-errors
aws s3 cp "${DEB_FILE}" "${R2_PREFIX}" --endpoint-url "$R2_ENDPOINT" --only-show-errors
# Stable "latest" alias so tests can fetch the newest nightly
# without knowing today's date.
echo "📤 Uploading latest alias"
"${AWS_CLI}" s3 cp "${DEB_FILE}" "${R2_PREFIX}rustfs-nightly-latest.deb" \
aws s3 cp "${DEB_FILE}" "${R2_PREFIX}rustfs-nightly-latest.deb" \
--endpoint-url "$R2_ENDPOINT" --only-show-errors
echo "✅ R2 upload complete"
CANDIDATE_FILE="${RUNNER_TEMP}/nightly-candidate-${GITHUB_RUN_ID}-${GITHUB_RUN_ATTEMPT}.json"
jq -n --arg source_sha "${SOURCE_SHA}" \
--argjson build_run_id "${GITHUB_RUN_ID}" --argjson build_run_attempt "${GITHUB_RUN_ATTEMPT}" \
--arg package_url "${CANDIDATE_URL}" --arg package_sha256 "${DEB_SHA256}" \
'{schema: 1, source_sha: $source_sha, build_run_id: $build_run_id, build_run_attempt: $build_run_attempt, package_url: $package_url, package_sha256: $package_sha256}' \
> "${CANDIDATE_FILE}"
echo "candidate_file=${CANDIDATE_FILE}" >> "${GITHUB_OUTPUT}"
- name: Upload nightly candidate manifest
if: ${{ steps.publish.outputs.candidate_file != '' }}
uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6
with:
name: nightly-candidate-${{ github.run_id }}-${{ github.run_attempt }}
path: ${{ steps.publish.outputs.candidate_file }}
if-no-files-found: error
# Live-Vault lane for the rustfs-kms suite (rustfs/backlog#1774).
#
# RUSTFS_KMS_VAULT_TOKEN is the single switch that adds the Vault KV2 and
+460 -44
View File
@@ -43,6 +43,10 @@ const ERR_LIFECYCLE_BUCKET_LOCKED: &str =
"ExpiredObjectAllVersions element and DelMarkerExpiration action cannot be used on an object locked bucket";
const ERR_LIFECYCLE_TOO_MANY_RULES: &str = "Lifecycle configuration should have at most 1000 rules";
const ERR_LIFECYCLE_INVALID_EXPIRATION_DAYS: &str = "'Days' for Expiration action must be a positive integer";
const ERR_LIFECYCLE_EXPIRATION_DAYS_DATE_CONFLICT: &str = "Expiration cannot specify both Days and Date";
const ERR_LIFECYCLE_MULTIPLE_TRANSITIONS: &str = "Only one Transition action per lifecycle rule is supported";
const ERR_LIFECYCLE_MULTIPLE_NONCURRENT_TRANSITIONS: &str =
"Only one NoncurrentVersionTransition action per lifecycle rule is supported";
const ERR_LIFECYCLE_INVALID_NONCURRENT_EXPIRATION_DAYS: &str =
"'NoncurrentDays' for NoncurrentVersionExpiration action must be a positive integer";
const ERR_LIFECYCLE_INVALID_ABORT_INCOMPLETE_MPU_DAYS: &str =
@@ -361,6 +365,12 @@ impl Lifecycle for BucketLifecycleConfiguration {
{
return Err(std::io::Error::other(ERR_LIFECYCLE_INVALID_EXPIRED_OBJECT_ALL_VERSIONS));
}
if expiration.days.is_some() && expiration.date.is_some() {
return Err(std::io::Error::new(
std::io::ErrorKind::InvalidInput,
ERR_LIFECYCLE_EXPIRATION_DAYS_DATE_CONFLICT,
));
}
if let Some(expiration_date) = &expiration.date {
let date = OffsetDateTime::from(expiration_date.clone());
if date.hour() != 0 || date.minute() != 0 || date.second() != 0 || date.nanosecond() != 0 {
@@ -394,11 +404,20 @@ impl Lifecycle for BucketLifecycleConfiguration {
}
}
if let Some(transitions) = &r.transitions {
if transitions.len() > 1 {
return Err(std::io::Error::new(std::io::ErrorKind::InvalidInput, ERR_LIFECYCLE_MULTIPLE_TRANSITIONS));
}
for transition in transitions {
TransitionOps::validate(transition)?;
}
}
if let Some(noncurrent_transitions) = &r.noncurrent_version_transitions {
if noncurrent_transitions.len() > 1 {
return Err(std::io::Error::new(
std::io::ErrorKind::InvalidInput,
ERR_LIFECYCLE_MULTIPLE_NONCURRENT_TRANSITIONS,
));
}
for transition in noncurrent_transitions {
NoncurrentVersionTransitionOps::validate(transition)?;
}
@@ -473,6 +492,8 @@ impl Lifecycle for BucketLifecycleConfiguration {
}
async fn eval(&self, obj: &ObjectOpts) -> Event {
// A single-object lookup cannot prove how many newer historical versions
// survive. Count-dependent actions wait for the complete-group evaluator.
self.eval_inner(obj, OffsetDateTime::now_utc(), 0).await
}
@@ -536,23 +557,8 @@ impl Lifecycle for BucketLifecycleConfiguration {
return Event::default();
};
if let Some(restore_expires) = obj.restore_expires
&& restore_expires.unix_timestamp() != 0
&& now.unix_timestamp() > restore_expires.unix_timestamp()
{
let mut action = IlmAction::DeleteRestoredAction;
if !obj.is_latest {
action = IlmAction::DeleteRestoredVersionAction;
}
events.push(Event {
action,
due: Some(now),
rule_id: "".into(),
noncurrent_days: 0,
newer_noncurrent_versions: 0,
storage_class: "".into(),
});
if let Some(event) = obj.restored_copy_expiry(now) {
events.push(event);
}
if let Some(ref lc_rules) = self.filter_rules(obj).await {
@@ -611,17 +617,12 @@ impl Lifecycle for BucketLifecycleConfiguration {
continue;
}
if !obj.is_latest
&& let Some(ref noncurrent_version_expiration) = rule.noncurrent_version_expiration
&& let Some(retain_newer_noncurrent_versions) = noncurrent_version_expiration.newer_noncurrent_versions
&& newer_noncurrent_versions < usize::try_from(retain_newer_noncurrent_versions).unwrap_or(usize::MAX)
{
continue;
}
if !obj.is_latest
&& let Some(ref noncurrent_version_expiration) = rule.noncurrent_version_expiration
&& let Some(noncurrent_days) = noncurrent_version_expiration.noncurrent_days
&& noncurrent_version_expiration
.newer_noncurrent_versions
.is_none_or(|retain| usize::try_from(retain).is_ok_and(|retain| newer_noncurrent_versions >= retain))
{
if let Some(successor_mod_time) = obj.successor_mod_time {
let expected_expiry = expected_expiry_time(successor_mod_time, noncurrent_days);
@@ -651,7 +652,11 @@ impl Lifecycle for BucketLifecycleConfiguration {
&& let Some(noncurrent_version_transition) = rule
.noncurrent_version_transitions
.as_ref()
.filter(|transitions| transitions.len() == 1)
.and_then(|transitions| transitions.first())
&& noncurrent_version_transition
.newer_noncurrent_versions
.is_none_or(|retain| usize::try_from(retain).is_ok_and(|retain| newer_noncurrent_versions >= retain))
&& let Some(storage_class) = noncurrent_version_transition.storage_class.as_ref()
&& !storage_class.as_str().is_empty()
&& !obj.delete_marker
@@ -735,7 +740,11 @@ impl Lifecycle for BucketLifecycleConfiguration {
}
if obj.transition_status != TRANSITION_COMPLETE
&& let Some(transition) = rule.transitions.as_ref().and_then(|transitions| transitions.first())
&& let Some(transition) = rule
.transitions
.as_ref()
.filter(|transitions| transitions.len() == 1)
.and_then(|transitions| transitions.first())
&& let Some(storage_class) = transition.storage_class.as_ref()
&& !storage_class.as_str().is_empty()
{
@@ -758,18 +767,15 @@ impl Lifecycle for BucketLifecycleConfiguration {
}
if !events.is_empty() {
// Select the winning event using a strict total order (MinIO semantics):
// the earliest `due` wins, and ties break toward delete-type actions. A
// missing `due` is treated as UNIX_EPOCH. This replaces a hand-written
// `sort_by` comparator that was not a strict weak ordering (it could return
// `Ordering::Less` for both `(a, b)` and `(b, a)`), which panics on the
// repository toolchain and did not deterministically pick the earliest event.
// Eligible expiration takes precedence over transition, even when a
// failed transition has an earlier deadline. Within each action class,
// prefer the earliest deadline using a deterministic total order.
let event = events
.iter()
.min_by_key(|event| {
(
event.due.unwrap_or(OffsetDateTime::UNIX_EPOCH).unix_timestamp(),
ilm_action_priority_rank(&event.action),
event.due.unwrap_or(OffsetDateTime::UNIX_EPOCH).unix_timestamp(),
)
})
.cloned()
@@ -1042,6 +1048,27 @@ impl ObjectOpts {
pub fn expired_object_deletemarker(&self) -> bool {
self.delete_marker && self.is_latest && self.num_versions == 1
}
pub(crate) fn restored_copy_expiry(&self, now: OffsetDateTime) -> Option<Event> {
let restore_expires = self.restore_expires?;
// Restore metadata alone does not prove that a durable remote copy exists.
if self.transition_status != TRANSITION_COMPLETE
|| restore_expires.unix_timestamp() == 0
|| now.unix_timestamp() <= restore_expires.unix_timestamp()
{
return None;
}
let action = if self.is_latest {
IlmAction::DeleteRestoredAction
} else {
IlmAction::DeleteRestoredVersionAction
};
expiration_action_has_valid_target(action, self.version_id, self.is_latest, self.delete_marker).then(|| Event {
action,
due: Some(now),
..Default::default()
})
}
}
/// Returns whether an expiry action has enough identity to target the object
@@ -1064,11 +1091,8 @@ pub fn expiration_action_has_valid_target(
}
}
/// Total-order rank for lifecycle actions used to break `due` ties.
///
/// Delete-type actions rank before every other action so that, when two events
/// share the same `due`, a delete wins (MinIO semantics). The concrete numeric
/// values only matter relative to each other.
/// Eligible logical expiration takes precedence over transition and restore-copy
/// cleanup. Deadlines break ties within an action class.
fn ilm_action_priority_rank(action: &IlmAction) -> u8 {
match action {
IlmAction::DeleteAllVersionsAction
@@ -4159,6 +4183,392 @@ mod tests {
assert_eq!(event.action, IlmAction::NoneAction);
}
mod adversarial_regressions {
use super::*;
use s3s::dto::NoncurrentVersionExpiration;
fn run(test: impl std::future::Future<Output = ()>) {
with_default_ilm_process_time(|| {
tokio::runtime::Builder::new_current_thread()
.build()
.expect("lifecycle regression runtime should build")
.block_on(test);
});
}
fn noncurrent_object() -> ObjectOpts {
ObjectOpts {
name: "logs/object".to_string(),
mod_time: Some(datetime!(2020-01-01 00:00:00 UTC)),
successor_mod_time: Some(datetime!(2020-01-02 00:00:00 UTC)),
version_id: Some(Uuid::from_u128(1)),
size: 1024 * 1024,
..Default::default()
}
}
#[test]
#[serial]
fn noncurrent_transition_retains_the_requested_newer_versions() {
run(async {
let mut rule = enabled_rule(None, None, Some("retain-two-hot-versions"));
rule.filter = Some(LifecycleRuleFilter::default());
rule.noncurrent_version_transitions = Some(vec![NoncurrentVersionTransition {
noncurrent_days: Some(1),
newer_noncurrent_versions: Some(2),
storage_class: Some(TransitionStorageClass::from_static("WARM")),
}]);
let lc = Arc::new(BucketLifecycleConfiguration {
rules: vec![rule],
expiry_updated_at: None,
});
lc.validate(&ObjectLockConfiguration::default())
.await
.expect("valid noncurrent transition policy");
let objects = (0..4)
.map(|index| ObjectOpts {
mod_time: Some(datetime!(2020-01-05 00:00:00 UTC) - Duration::days(index)),
successor_mod_time: (index > 0).then_some(datetime!(2020-01-06 00:00:00 UTC) - Duration::days(index)),
version_id: Some(Uuid::from_u128(u128::try_from(index + 1).expect("small version index"))),
is_latest: index == 0,
num_versions: 4,
..noncurrent_object()
})
.collect::<Vec<_>>();
let actions = crate::Evaluator::new(lc)
.eval(&objects)
.await
.expect("complete version chain should evaluate")
.into_iter()
.map(|event| event.action)
.collect::<Vec<_>>();
assert_eq!(
actions,
[
IlmAction::NoneAction,
IlmAction::NoneAction,
IlmAction::NoneAction,
IlmAction::TransitionVersionAction
],
"the two newest noncurrent versions must remain in their current storage class"
);
});
}
#[test]
#[serial]
fn noncurrent_transition_checks_count_age_and_single_object_context() {
run(async {
let mut rule = enabled_rule(None, None, Some("retain-two"));
rule.filter = Some(LifecycleRuleFilter::default());
rule.noncurrent_version_transitions = Some(vec![NoncurrentVersionTransition {
noncurrent_days: Some(3),
newer_noncurrent_versions: Some(2),
storage_class: Some(TransitionStorageClass::from_static("WARM")),
}]);
let mut lc = BucketLifecycleConfiguration {
rules: vec![rule],
expiry_updated_at: None,
};
lc.validate(&ObjectLockConfiguration::default())
.await
.expect("valid counted transition");
let object = noncurrent_object();
let now = datetime!(2020-01-10 00:00:00 UTC);
for (newer, expected) in [
(0, IlmAction::NoneAction),
(1, IlmAction::NoneAction),
(2, IlmAction::TransitionVersionAction),
(3, IlmAction::TransitionVersionAction),
] {
assert_eq!(lc.eval_inner(&object, now, newer).await.action, expected, "newer count: {newer}");
}
assert_eq!(
lc.eval_inner(&object, datetime!(2020-01-04 00:00:00 UTC), 2).await.action,
IlmAction::NoneAction,
"the retention count does not replace the age condition"
);
assert_eq!(
lc.eval(&object).await.action,
IlmAction::NoneAction,
"a single-object lookup must not assume a complete version history"
);
for retain in [None, Some(0), Some(-1), Some(i32::MAX)] {
lc.rules[0]
.noncurrent_version_transitions
.as_mut()
.expect("transition exists")[0]
.newer_noncurrent_versions = retain;
let expected = if matches!(retain, None | Some(0)) {
IlmAction::TransitionVersionAction
} else {
IlmAction::NoneAction
};
assert_eq!(lc.eval_inner(&object, now, 2).await.action, expected, "retention: {retain:?}");
}
});
}
#[test]
#[serial]
fn noncurrent_expiration_and_transition_have_independent_retention_counts() {
run(async {
let mut rule = enabled_rule(None, None, Some("independent-counts"));
rule.filter = Some(LifecycleRuleFilter::default());
rule.noncurrent_version_expiration = Some(NoncurrentVersionExpiration {
noncurrent_days: Some(90),
newer_noncurrent_versions: Some(4),
});
rule.noncurrent_version_transitions = Some(vec![NoncurrentVersionTransition {
noncurrent_days: Some(30),
newer_noncurrent_versions: Some(2),
storage_class: Some(TransitionStorageClass::from_static("WARM")),
}]);
let lc = BucketLifecycleConfiguration {
rules: vec![rule],
expiry_updated_at: None,
};
lc.validate(&ObjectLockConfiguration::default())
.await
.expect("valid independent retention limits");
let object = noncurrent_object();
let now = datetime!(2020-05-01 00:00:00 UTC);
for (newer, expected) in [
(1, IlmAction::NoneAction),
(2, IlmAction::TransitionVersionAction),
(3, IlmAction::TransitionVersionAction),
(4, IlmAction::DeleteVersionAction),
] {
assert_eq!(lc.eval_inner(&object, now, newer).await.action, expected, "newer count: {newer}");
}
});
}
#[test]
#[serial]
fn expiration_retention_does_not_skip_an_independent_transition() {
run(async {
let mut rule = enabled_rule(None, None, Some("transition-then-expire"));
rule.filter = Some(LifecycleRuleFilter::default());
rule.noncurrent_version_transitions = Some(vec![NoncurrentVersionTransition {
noncurrent_days: Some(1),
newer_noncurrent_versions: None,
storage_class: Some(TransitionStorageClass::from_static("WARM")),
}]);
let mut lc = BucketLifecycleConfiguration {
rules: vec![rule],
expiry_updated_at: None,
};
let object = noncurrent_object();
let now = datetime!(2020-01-10 00:00:00 UTC);
let transition_only = lc.eval_inner(&object, now, 0).await;
assert_eq!(transition_only.action, IlmAction::TransitionVersionAction);
lc.rules[0].noncurrent_version_expiration = Some(NoncurrentVersionExpiration {
noncurrent_days: Some(90),
newer_noncurrent_versions: Some(2),
});
lc.validate(&ObjectLockConfiguration::default())
.await
.expect("valid combined policy");
let combined = lc.eval_inner(&object, now, 0).await;
assert_eq!(combined.action, transition_only.action, "retention limits expiration, not transition");
assert_eq!(combined.storage_class, transition_only.storage_class);
});
}
#[test]
#[serial]
fn current_transition_rejects_multiple_stages_in_any_order() {
run(async {
let mut rule = enabled_rule(None, None, Some("two-current-transitions"));
rule.transitions = Some(vec![
Transition {
date: Some(datetime!(2020-03-01 00:00:00 UTC).into()),
days: None,
storage_class: Some(TransitionStorageClass::from_static("COLD")),
},
Transition {
date: Some(datetime!(2020-01-03 00:00:00 UTC).into()),
days: None,
storage_class: Some(TransitionStorageClass::from_static("WARM")),
},
]);
let mut lc = BucketLifecycleConfiguration {
rules: vec![rule],
expiry_updated_at: None,
};
let object = ObjectOpts {
is_latest: true,
..noncurrent_object()
};
let now = datetime!(2020-01-10 00:00:00 UTC);
for status in [ExpirationStatus::ENABLED, ExpirationStatus::DISABLED] {
lc.rules[0].status = ExpirationStatus::from_static(status);
for _ in 0..2 {
let err = lc
.validate(&ObjectLockConfiguration::default())
.await
.expect_err("multiple transition stages must be rejected");
assert_eq!(err.kind(), std::io::ErrorKind::InvalidInput);
assert_eq!(err.to_string(), ERR_LIFECYCLE_MULTIPLE_TRANSITIONS);
assert_eq!(
lc.eval_inner(&object, now, 0).await.action,
IlmAction::NoneAction,
"legacy multi-stage configurations must not silently execute their first stage"
);
lc.rules[0]
.transitions
.as_mut()
.expect("transition array is present")
.reverse();
}
}
lc.rules[0]
.transitions
.as_mut()
.expect("transition array is present")
.remove(0);
lc.rules[0].status = ExpirationStatus::from_static(ExpirationStatus::ENABLED);
lc.validate(&ObjectLockConfiguration::default())
.await
.expect("one stage is supported");
let event = lc.eval_inner(&object, now, 0).await;
assert_eq!(event.action, IlmAction::TransitionAction);
assert_eq!(event.storage_class, "WARM");
});
}
#[test]
#[serial]
fn noncurrent_transition_rejects_multiple_stages_in_any_order() {
run(async {
let mut rule = enabled_rule(None, None, Some("two-noncurrent-transitions"));
rule.noncurrent_version_transitions = Some(vec![
NoncurrentVersionTransition {
noncurrent_days: Some(30),
newer_noncurrent_versions: None,
storage_class: Some(TransitionStorageClass::from_static("COLD")),
},
NoncurrentVersionTransition {
noncurrent_days: Some(1),
newer_noncurrent_versions: None,
storage_class: Some(TransitionStorageClass::from_static("WARM")),
},
]);
let mut lc = BucketLifecycleConfiguration {
rules: vec![rule],
expiry_updated_at: None,
};
let object = noncurrent_object();
let now = datetime!(2020-01-10 00:00:00 UTC);
for status in [ExpirationStatus::ENABLED, ExpirationStatus::DISABLED] {
lc.rules[0].status = ExpirationStatus::from_static(status);
for _ in 0..2 {
let err = lc
.validate(&ObjectLockConfiguration::default())
.await
.expect_err("multiple noncurrent transition stages must be rejected");
assert_eq!(err.kind(), std::io::ErrorKind::InvalidInput);
assert_eq!(err.to_string(), ERR_LIFECYCLE_MULTIPLE_NONCURRENT_TRANSITIONS);
assert_eq!(
lc.eval_inner(&object, now, 0).await.action,
IlmAction::NoneAction,
"legacy multi-stage configurations must not silently execute their first stage"
);
lc.rules[0]
.noncurrent_version_transitions
.as_mut()
.expect("transition array is present")
.reverse();
}
}
lc.rules[0]
.noncurrent_version_transitions
.as_mut()
.expect("transition array is present")
.remove(0);
lc.rules[0].status = ExpirationStatus::from_static(ExpirationStatus::ENABLED);
lc.validate(&ObjectLockConfiguration::default())
.await
.expect("one stage is supported");
let event = lc.eval_inner(&object, now, 0).await;
assert_eq!(event.action, IlmAction::TransitionVersionAction);
assert_eq!(event.storage_class, "WARM");
});
}
#[test]
#[serial]
fn expiration_rejects_simultaneous_days_and_date() {
run(async {
let mut lc = BucketLifecycleConfiguration {
rules: vec![enabled_rule(
Some(LifecycleExpiration {
days: Some(1),
..Default::default()
}),
None,
Some("ambiguous-expiry"),
)],
expiry_updated_at: None,
};
lc.validate(&ObjectLockConfiguration::default())
.await
.expect("a single Days expiration is valid");
lc.rules[0].expiration.as_mut().expect("expiration is present").date =
Some(datetime!(2099-01-01 00:00:00 UTC).into());
let err = lc
.validate(&ObjectLockConfiguration::default())
.await
.expect_err("Days and Date are mutually exclusive; accepting both silently overrides Days");
assert_eq!(err.kind(), std::io::ErrorKind::InvalidInput);
assert_eq!(err.to_string(), ERR_LIFECYCLE_EXPIRATION_DAYS_DATE_CONFLICT);
});
}
#[test]
#[serial]
fn overdue_transition_does_not_starve_permanent_expiration() {
run(async {
let mut rule = enabled_rule(
Some(LifecycleExpiration {
days: Some(90),
..Default::default()
}),
None,
Some("archive-then-delete"),
);
rule.transitions = Some(vec![Transition {
days: Some(30),
date: None,
storage_class: Some(TransitionStorageClass::from_static("WARM")),
}]);
let lc = BucketLifecycleConfiguration {
rules: vec![rule],
expiry_updated_at: None,
};
lc.validate(&ObjectLockConfiguration::default())
.await
.expect("valid transition and expiration policy");
let object = ObjectOpts {
is_latest: true,
version_id: None,
transition_status: TRANSITION_PENDING.to_string(),
..noncurrent_object()
};
let before_expiration = lc.eval_inner(&object, datetime!(2020-02-15 00:00:00 UTC), 0).await;
assert_eq!(before_expiration.action, IlmAction::TransitionAction);
let overdue = lc.eval_inner(&object, datetime!(2020-05-01 00:00:00 UTC), 0).await;
assert_eq!(
overdue.action,
IlmAction::DeleteAction,
"an unavailable tier must not prevent permanent expiration indefinitely"
);
});
}
}
/// Property-based tests for the rule evaluator (backlog#1148 ilm-14,
/// follow-up to backlog#1030 / rustfs#4455).
///
@@ -4169,7 +4579,7 @@ mod tests {
///
/// * `eval_inner` never panics and is deterministic for a fixed input;
/// * the winning event matches an independently recomputed candidate set:
/// earliest `due` wins, ties break toward delete-class actions (the
/// eligible expiration wins over transition, then earliest `due` wins (the
/// `min_by_key` selection that replaced the rustfs#4455 comparator);
/// * `expected_expiry_time` is monotonically non-decreasing in `days` and
/// always lands on the processing boundary, both at production defaults
@@ -4458,8 +4868,8 @@ mod tests {
/// consider for a live current version under `selection`-shaped rules
/// (expiration and first-transition only, no filters): expiration
/// fires when `now >= due`, transition when `now > due` and the object
/// has not already transitioned. Selection semantics under test:
/// earliest due wins, ties prefer delete-class.
/// has not already transitioned. Eligible expiration wins over transition;
/// the earliest deadline wins within the selected action class.
fn oracle_candidates(lc: &BucketLifecycleConfiguration, obj: &ObjectOpts, now: OffsetDateTime) -> Vec<Candidate> {
let mod_time = obj.mod_time.expect("selection strategy always sets mod_time");
let mut candidates = Vec::new();
@@ -4548,8 +4958,8 @@ mod tests {
/// Differential test of winner selection (the rustfs#4455 fix):
/// for a live current version under randomized expiration and
/// transition rules, `eval_inner`'s winner must carry the
/// minimum `(due, rank)` of the independently recomputed
/// candidate set — earliest due wins, ties prefer delete-class —
/// earliest expiration from the independently recomputed candidate
/// set, or the earliest transition when no expiration is eligible,
/// and must be `NoneAction` exactly when that set is empty.
#[test]
#[serial]
@@ -4578,7 +4988,13 @@ mod tests {
// Oracle and evaluator must observe the same (pinned) time env.
let (event, expected) = with_production_time_env(|| {
let expected = oracle_candidates(&lc, &obj, now).into_iter().min();
let candidates = oracle_candidates(&lc, &obj, now);
let expected = candidates
.iter()
.filter(|(_, rank)| *rank == 0)
.min()
.copied()
.or_else(|| candidates.into_iter().min());
let rt = tokio::runtime::Builder::new_current_thread()
.enable_all()
.build()
+93 -7
View File
@@ -116,13 +116,10 @@ impl Evaluator {
break 'top_loop;
}
}
IlmAction::DeleteAction
| IlmAction::DeleteRestoredAction
| IlmAction::DeleteVersionAction
| IlmAction::DeleteRestoredVersionAction
if self.is_object_locked(obj) =>
{
event = Event::default();
// Restore expiry removes only the temporary local copy; the
// retained logical version and its remote data remain intact.
IlmAction::DeleteAction | IlmAction::DeleteVersionAction if self.is_object_locked(obj) => {
event = obj.restored_copy_expiry(now).unwrap_or_default();
}
_ => {}
}
@@ -206,6 +203,95 @@ mod tests {
use super::*;
use rustfs_replication::{ReplicationStatusType, VersionPurgeStatusType};
#[tokio::test]
async fn adversarial_restore_expiry_survives_legal_hold() {
let mut policy = (*latest_expiration_lifecycle()).clone();
policy.rules[0].status = ExpirationStatus::from_static(ExpirationStatus::DISABLED);
let policy = Arc::new(policy);
policy
.validate(&lock_enabled_without_default_retention())
.await
.expect("valid disabled lifecycle rule");
let mut objects = [true, false].map(|is_latest| ObjectOpts {
is_latest,
num_versions: 2,
mod_time: Some(
OffsetDateTime::from_unix_timestamp(if is_latest { 1_200_000 } else { 1_000_000 })
.expect("fixed version timestamp"),
),
successor_mod_time: (!is_latest)
.then(|| OffsetDateTime::from_unix_timestamp(1_200_000).expect("fixed successor timestamp")),
transition_status: crate::TRANSITION_COMPLETE.to_string(),
restore_expires: Some(OffsetDateTime::from_unix_timestamp(2_000_000).expect("fixed expired restore timestamp")),
..current_object_opts(ReplicationStatusType::Completed)
});
let evaluator = Evaluator::new(policy).with_lock_retention(Some(lock_enabled_without_default_retention()));
let expected = [IlmAction::DeleteRestoredAction, IlmAction::DeleteRestoredVersionAction];
let unlocked = evaluator
.eval(&objects)
.await
.expect("unlocked restored versions should evaluate");
assert_eq!(unlocked.iter().map(|event| event.action).collect::<Vec<_>>(), expected);
for object in &mut objects {
object
.user_defined
.insert(X_AMZ_OBJECT_LOCK_LEGAL_HOLD.as_str().to_string(), "ON".to_string());
}
let locked = evaluator
.eval(&objects)
.await
.expect("locked restored versions should evaluate");
assert_eq!(
locked.iter().map(|event| event.action).collect::<Vec<_>>(),
expected,
"expiring a restored local copy preserves the retained logical version and remote object"
);
let mut expiring_policy = (*latest_expiration_lifecycle()).clone();
expiring_policy.rules[0].noncurrent_version_expiration = Some(NoncurrentVersionExpiration {
noncurrent_days: Some(1),
newer_noncurrent_versions: None,
});
let expiring_evaluator =
Evaluator::new(Arc::new(expiring_policy)).with_lock_retention(Some(lock_enabled_without_default_retention()));
let locked = expiring_evaluator
.eval(&objects)
.await
.expect("locked expired versions should evaluate");
assert_eq!(
locked.iter().map(|event| event.action).collect::<Vec<_>>(),
expected,
"blocked logical expiration must still allow an eligible restore-copy cleanup"
);
for status in [ReplicationStatusType::Pending, ReplicationStatusType::Failed] {
for object in &mut objects {
object.replication_status = status.clone();
}
for evaluator in [&evaluator, &expiring_evaluator] {
let events = evaluator.eval(&objects).await.expect("pending replication should evaluate");
assert!(events.iter().all(|event| event.action == IlmAction::NoneAction));
}
}
for object in &mut objects {
object.replication_status = ReplicationStatusType::Completed;
}
for transition_status in ["", crate::TRANSITION_PENDING, "unknown"] {
for object in &mut objects {
object.transition_status = transition_status.to_string();
}
for evaluator in [&evaluator, &expiring_evaluator] {
let events = evaluator.eval(&objects).await.expect("incomplete transition should evaluate");
assert!(
events.iter().all(|event| event.action == IlmAction::NoneAction),
"restore metadata cannot authorize cleanup without a completed transition"
);
}
}
}
fn expired_marker_lifecycle() -> Arc<BucketLifecycleConfiguration> {
Arc::new(BucketLifecycleConfiguration {
expiry_updated_at: None,
+10
View File
@@ -22,6 +22,16 @@
| `FileMeta` / `FileInfo` / version metadata | `crates/filemeta/src/` |
| Dual-key internal metadata helpers (`insert_bytes` / `get_bytes`) | `crates/utils/src/http/metadata_compat.rs` |
## Lifecycle rule limits and evaluation
Each lifecycle rule supports at most one `Transition` and one `NoncurrentVersionTransition`. A version can make one initial transition; chaining additional tiers after it reaches `complete` is not supported. Splitting stages across overlapping rules does not enable a transition chain. `PutBucketLifecycleConfiguration` rejects multiple entries in either transition array with `InvalidArgument`, including in disabled rules. Existing stored multi-entry arrays are not executed; replace each with a single intended destination. Independent expiration actions in the rule remain eligible.
`Expiration.Days` and `Expiration.Date` are mutually exclusive. A request containing both is rejected instead of silently selecting the date. When expiration and transition are both eligible, expiration takes precedence; a failed earlier transition does not keep an expired object indefinitely. Deadlines select the earliest action within the same action class.
Noncurrent expiration and transition have independent `NewerNoncurrentVersions` limits. A transition with a positive limit waits for a complete version-group evaluation to establish that enough newer noncurrent versions remain. Single-object evaluation, including the current manual transition and immediate-enqueue paths, conservatively defers these counted transitions to the lifecycle scanner. An unmet expiration retention limit does not suppress a separately eligible transition.
An expired restored local copy can be cleaned up under Object Lock because the retained logical version and remote data remain intact. Cleanup requires a completed transition and still waits for pending or failed replication. The storage layer revalidates the source identity and restore metadata before removing the local copy; restore headers alone do not authorize cleanup.
## Free-version recovery controls
The dedicated free-version recovery loop is enabled by default and is independent of the data scanner and heal switches. Setting `RUSTFS_SCANNER_ENABLED=false` does not stop this repair loop. Set `RUSTFS_TIER_FREE_VERSION_RECOVERY_ENABLED=false` before process startup to disable only the dedicated persisted-marker walk. That setting does not disable lifecycle workers or prevent another scanner path from discovering a free version, and it can leave remote cleanup markers pending for longer, so use it as a break-glass pressure control rather than a cleanup mechanism.
-208
View File
@@ -1,208 +0,0 @@
#!/usr/bin/env python3
"""Exercise the nightly publication step without AWS, network or package builds."""
import hashlib
import json
import os
from pathlib import Path
import subprocess
import tempfile
import unittest
from check_test_wiring import yaml_block
ROOT = Path(__file__).resolve().parents[1]
WORKFLOW = ROOT / ".github/workflows/nightly-gnu.yml"
class NightlyCandidateTests(unittest.TestCase):
def setUp(self):
self.temp = tempfile.TemporaryDirectory()
self.addCleanup(self.temp.cleanup)
self.root = Path(self.temp.name)
self.package = self.root / "rustfs-nightly-2026-09-06.deb"
self.package.write_bytes(b"built package bytes\x00\xff")
for command in (["git", "init", "-q"], ["git", "add", self.package.name],
["git", "-c", "user.name=Test", "-c", "user.email=test@example.invalid", "commit", "-qm", "fixture"]):
subprocess.run(command, cwd=self.root, check=True, capture_output=True)
self.sha = subprocess.check_output(["git", "rev-parse", "HEAD"], cwd=self.root, text=True).strip()
self.digest = hashlib.sha256(self.package.read_bytes()).hexdigest()
self.output = self.root / "github-output"
self.store = self.root / "store"
self.shims = self.root / "fake-tools.sh"
self.shims.write_text(r'''aws() {
printf '%s\n' "$*" >> "$FAKE_AWS_LOG"
if [[ "$1" == --version ]]; then printf 'aws-cli/1.44.79 fixture\n'; return; fi
if [[ "$*" == *--generate-cli-skeleton* ]]; then
if [[ "$FAKE_MODE" == broken-install || "$FAKE_MODE" =~ ^(old-cli|bootstrap-failure|install-failure)$ && ! -e "$FAKE_INSTALLED" ]]; then
printf '{}\n'
else
printf '{"IfNoneMatch":""}\n'
fi
return
fi
if [[ "$1 $2" == 's3api put-object' ]]; then
[[ "$FAKE_MODE" != upload-failure ]] || return 42
shift 2
local key="" body="" condition=""
while [[ $# -gt 0 ]]; do
case "$1" in
--key) key="$2";;
--body) body="$2";;
--if-none-match) condition="$2";;
esac
shift 2
done
[[ -z "$condition" || "$condition" == '*' ]] || return 43
if [[ "$condition" == '*' && -e "$FAKE_STORE/$key" ]]; then return 44; fi
mkdir -p "$(dirname "$FAKE_STORE/$key")"
cp "$body" "$FAKE_STORE/$key"
elif [[ "$1 $2" == 's3 cp' ]]; then
[[ "$FAKE_MODE" != alias-failure ]] || return 45
local destination="${4#s3://test-bucket/}"
[[ "$destination" != */ ]] || destination+="$(basename "$3")"
mkdir -p "$(dirname "$FAKE_STORE/$destination")"
cp "$3" "$FAKE_STORE/$destination"
else
return 46
fi
}
curl() {
local url="${!#}"
printf '%s\n' "$url" >> "$FAKE_CURL_LOG"
[[ "$url" == https://dl.rustfs.com/artifacts/rustfs/packages/nightly/runs/* ]] || return 22
[[ "$FAKE_MODE" != missing-public-url ]] || return 22
if [[ "$FAKE_MODE" == wrong-public-bytes ]]; then printf 'different package'; return; fi
cat "$FAKE_STORE/${url#https://dl.rustfs.com/}" || return 22
[[ "$FAKE_MODE" != incomplete-download ]] || return 47
}
sudo() {
[[ "$*" == 'apt-get update' || "$*" == 'apt-get install -y -qq python3-venv' ]] || return 49
[[ "$FAKE_MODE" != bootstrap-failure ]] || return 48
}
python3() {
[[ "$1 $2" == '-m venv' ]] || return 50
mkdir -p "$3/bin"
cat > "$3/bin/python" <<'SH'
#!/usr/bin/env bash
[[ "$*" == '-m pip install --disable-pip-version-check awscli==1.44.79' ]] || exit 51
[[ "$FAKE_MODE" != install-failure ]] || exit 52
: > "$FAKE_INSTALLED"
SH
printf '#!/usr/bin/env bash\naws "$@"\n' > "$3/bin/aws"
chmod +x "$3/bin/python" "$3/bin/aws"
}
''')
self.env = dict(os.environ, BASH_ENV=str(self.shims), DEB_FILE=self.package.name,
R2_ACCESS_KEY_ID="fake-access", R2_SECRET_ACCESS_KEY="fake-secret", R2_ENDPOINT="https://r2.example.invalid", R2_BUCKET="test-bucket",
RUNNER_TEMP=str(self.root), GITHUB_SHA=self.sha, GITHUB_RUN_ID="12345", GITHUB_RUN_ATTEMPT="1", GITHUB_OUTPUT=str(self.output),
FAKE_STORE=str(self.store), FAKE_AWS_LOG=str(self.root / "aws.log"), FAKE_CURL_LOG=str(self.root / "curl.log"), FAKE_INSTALLED=str(self.root / "installed"), FAKE_MODE="success")
source = WORKFLOW.read_text()
job = yaml_block(source.splitlines(), "build", 2)
starts = [i for i, line in enumerate(job) if line.startswith(" - name: ")]
self.steps = {
job[start].split(": ", 1)[1]: job[start:end]
for start, end in zip(starts, starts[1:] + [len(job)])
}
self.publish = self.steps["Upload DEB to Cloudflare R2"]
start = self.publish.index(" run: |") + 1
self.shell = "\n".join(line[10:] for line in self.publish[start:] if not line.strip() or line.startswith(" "))
def run_publish(self, **overrides):
self.output.unlink(missing_ok=True)
return subprocess.run(["bash", "--noprofile", "--norc", "-e", "-o", "pipefail", "-c", self.shell],
cwd=self.root, env=dict(self.env, **overrides), capture_output=True, text=True)
def manifest(self):
output = self.output.read_text().strip()
self.assertTrue(output.startswith("candidate_file="), output)
return json.loads(Path(output.split("=", 1)[1]).read_text())
def test_success_binds_actual_package_checkout_and_attempt(self):
result = self.run_publish()
self.assertEqual(result.returncode, 0, result.stderr)
manifest = self.manifest()
self.assertEqual(manifest, {"schema": 1, "source_sha": self.sha, "build_run_id": 12345, "build_run_attempt": 1,
"package_sha256": self.digest, "package_url": f"https://dl.rustfs.com/artifacts/rustfs/packages/nightly/runs/12345/1/{self.digest}/rustfs.deb"})
for path in (f"runs/12345/1/{self.digest}/rustfs.deb", self.package.name, "rustfs-nightly-latest.deb"):
self.assertEqual((self.store / "artifacts/rustfs/packages/nightly" / path).read_bytes(), self.package.read_bytes())
self.assertEqual((self.root / "curl.log").read_text().strip(), manifest["package_url"])
def test_missing_credentials_remain_artifact_only(self):
for key in ("R2_ACCESS_KEY_ID", "R2_SECRET_ACCESS_KEY", "R2_ENDPOINT", "R2_BUCKET"):
with self.subTest(missing=key):
result = self.run_publish(**{key: ""})
self.assertEqual(result.returncode, 0, result.stderr)
self.assertFalse(self.output.exists())
self.assertFalse((self.root / "aws.log").exists())
self.assertEqual(list(self.root.glob("nightly-candidate-*.json")), [])
def test_publication_failures_never_emit_a_candidate(self):
for index, mode in enumerate(("upload-failure", "missing-public-url", "wrong-public-bytes", "incomplete-download", "alias-failure")):
with self.subTest(mode=mode):
result = self.run_publish(FAKE_MODE=mode, GITHUB_RUN_ID=str(20000 + index))
self.assertNotEqual(result.returncode, 0, result.stdout)
self.assertFalse(self.output.exists())
self.assertEqual(list(self.root.glob("nightly-candidate-*.json")), [])
def test_old_cli_is_upgraded_in_an_isolated_temporary_environment(self):
result = self.run_publish(FAKE_MODE="old-cli")
self.assertEqual(result.returncode, 0, result.stderr)
self.assertTrue((self.root / "installed").exists())
self.assertEqual(self.manifest()["package_sha256"], self.digest)
self.assertEqual(list(self.root.glob("nightly-awscli.*")), [])
def test_failed_cli_bootstrap_cannot_publish(self):
for mode in ("bootstrap-failure", "install-failure", "broken-install"):
with self.subTest(mode=mode):
(self.root / "installed").unlink(missing_ok=True)
result = self.run_publish(FAKE_MODE=mode)
self.assertNotEqual(result.returncode, 0)
self.assertFalse(self.output.exists())
self.assertFalse(self.store.exists())
self.assertEqual(list(self.root.glob("nightly-awscli.*")), [])
def test_checkout_sha_mismatch_fails_before_upload(self):
result = self.run_publish(GITHUB_SHA="f" * 40)
self.assertNotEqual(result.returncode, 0)
self.assertIn("Checkout SHA", result.stderr)
self.assertFalse(self.output.exists())
self.assertFalse((self.root / "aws.log").exists())
def test_same_date_builds_and_reruns_keep_distinct_candidates(self):
urls = []
for run, attempt in (("12345", "1"), ("54321", "1"), ("12345", "2")):
result = self.run_publish(GITHUB_RUN_ID=run, GITHUB_RUN_ATTEMPT=attempt)
self.assertEqual(result.returncode, 0, result.stderr)
urls.append(self.manifest()["package_url"])
self.assertEqual(len(set(urls)), 3)
self.assertEqual(len(list(self.root.glob("nightly-candidate-*.json"))), 3)
def test_duplicate_key_is_not_overwritten_or_recertified(self):
result = self.run_publish()
self.assertEqual(result.returncode, 0, result.stderr)
key = self.manifest()["package_url"].removeprefix("https://dl.rustfs.com/")
stored = self.store / key
stored.write_bytes(b"preexisting conflicting object")
(self.root / "nightly-candidate-12345-1.json").unlink()
result = self.run_publish()
self.assertNotEqual(result.returncode, 0)
self.assertEqual(stored.read_bytes(), b"preexisting conflicting object")
self.assertFalse(self.output.exists())
self.assertEqual(list(self.root.glob("nightly-candidate-*.json")), [])
def test_manifest_upload_requires_publication_output(self):
upload = self.steps["Upload nightly candidate manifest"]
self.assertIn(" id: publish", self.publish)
self.assertIn(" DEB_FILE: ${{ steps.deb.outputs.deb_file }}", self.publish)
self.assertIn(" if: ${{ steps.publish.outputs.candidate_file != '' }}", upload)
self.assertIn(" name: nightly-candidate-${{ github.run_id }}-${{ github.run_attempt }}", upload)
self.assertIn(" path: ${{ steps.publish.outputs.candidate_file }}", upload)
self.assertIn(" if-no-files-found: error", upload)
self.assertNotIn(" continue-on-error: true", self.publish)
self.assertNotIn(" overwrite: true", upload)
if __name__ == "__main__":
unittest.main()