Compare commits

..

2 Commits

Author SHA1 Message Date
houseme b3c79cd71a Merge branch 'main' into overtrue/wait-bucket-metadata-reload 2026-08-23 12:30:36 +08:00
overtrue c57f22c3a0 fix(app): wait for peer bucket metadata reload 2026-08-22 15:49:37 +08:00
12 changed files with 172 additions and 478 deletions
-20
View File
@@ -1,20 +0,0 @@
# Report-only calibration baseline from https://github.com/rustfs/rustfs/actions/runs/29394996173.
# Update counts only with a linked coverage run and a reviewed explanation.
phase = "report-only"
allowed_drop_percentage_points = 1.0
[crates."crates/iam"]
covered = 5149
count = 8131
[crates."crates/kms"]
covered = 2950
count = 4200
[crates."crates/policy"]
covered = 4636
count = 5464
[crates."crates/crypto"]
covered = 469
count = 494
-1
View File
@@ -36,7 +36,6 @@ script-tests: ## Run shell script tests
./scripts/test_manual_transition_runbooks.sh
./scripts/check_embedded_secrets.sh --self-test
python3 ./scripts/check_test_wiring.py --self-test
python3 ./scripts/check_security_coverage.py --self-test
python3 ./scripts/check_scheduled_validation_freshness.py --self-test
python3 ./scripts/s3-tests/test_report_compat.py
bash -n ./scripts/validate_object_data_cache_cold_stampede.sh
+12 -28
View File
@@ -12,12 +12,14 @@
# See the License for the specific language governing permissions and
# limitations under the License.
# Workspace line-coverage baseline and security-crate calibration
# (backlog#1153 infra-5/infra-6).
# Weekly workspace line-coverage baseline (backlog#1153 infra-5).
#
# NON-BLOCKING by design: the weekly job gives coverage a visible baseline and
# trend, while relevant pull requests run a report-only security-crate
# comparison. Neither job is a required check during calibration.
# NON-BLOCKING by design: this workflow only runs on schedule and manual
# dispatch, so it never attaches a status to a PR and must never be made a
# required check. It exists to give coverage a visible baseline and trend
# (per-crate table in the job summary, lcov artifact kept 90 days) — the
# per-crate ratchet for the security-critical crates builds on it later
# (backlog#1153 infra-6, report-only first per the ci-11 ladder).
#
# Measurement scope matches the PR test gate (ci.yml "Run tests"):
# `--workspace --exclude e2e_test` with the `ci` nextest profile. Doctests are
@@ -29,17 +31,6 @@
name: coverage
on:
pull_request:
branches: [main]
paths:
- "crates/iam/**"
- "crates/kms/**"
- "crates/policy/**"
- "crates/crypto/**"
- ".config/coverage-baselines.toml"
- "scripts/coverage_per_crate.py"
- "scripts/check_security_coverage.py"
- ".github/workflows/coverage.yml"
workflow_dispatch:
schedule:
# 07:00 UTC Sunday — staggered clear of the other Sunday crons: ci (00:00),
@@ -48,10 +39,6 @@ on:
# e2e-replication-nightly (04:00) and performance-ab (06:00) lanes.
- cron: "43 7 * * 0"
concurrency:
group: ${{ github.workflow }}-${{ github.event_name }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: ${{ github.event_name != 'schedule' }}
# Only alert-on-failure needs more than read access; it declares its own
# job-level `issues: write`.
permissions:
@@ -59,13 +46,12 @@ permissions:
jobs:
coverage:
name: Workspace line coverage
name: Workspace coverage (weekly)
runs-on: sm-standard-4
# The instrumented build cannot reuse the regular CI cache (different
# RUSTFLAGS), so a cold run rebuilds the workspace before running the
# full suite. Exact-head run 32573798257 needed 119m42s including reports
# and artifact upload, so keep a bounded 30-minute publication margin.
timeout-minutes: 150
# RUSTFLAGS), so a cold week rebuilds the workspace before running the
# full suite; give it double the test job's 60-minute budget.
timeout-minutes: 120
env:
FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
# Match the PR gate's nextest semantics (ci.yml runs `--profile ci`):
@@ -105,9 +91,7 @@ jobs:
cargo llvm-cov report --json --output-path target/llvm-cov/coverage.json
- name: Write per-crate summary
run: |
python3 scripts/coverage_per_crate.py target/llvm-cov/coverage.json >> "$GITHUB_STEP_SUMMARY"
python3 scripts/check_security_coverage.py target/llvm-cov/coverage.json >> "$GITHUB_STEP_SUMMARY"
run: python3 scripts/coverage_per_crate.py target/llvm-cov/coverage.json >> "$GITHUB_STEP_SUMMARY"
- name: Upload coverage artifact
if: always()
+6 -3
View File
@@ -39,10 +39,11 @@ jobs:
env:
FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
steps:
- name: Checkout repository
- name: Checkout main branch
uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
with:
persist-credentials: false
ref: main
- name: Setup Rust environment
uses: ./.github/actions/setup
@@ -88,10 +89,11 @@ jobs:
# either casing.
NO_PROXY: 127.0.0.1,localhost
steps:
- name: Checkout repository
- name: Checkout main branch
uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
with:
persist-credentials: false
ref: main
- name: Setup Rust environment
uses: ./.github/actions/setup
@@ -176,10 +178,11 @@ jobs:
FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
NO_PROXY: 127.0.0.1,localhost
steps:
- name: Checkout repository
- name: Checkout main branch
uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
with:
persist-credentials: false
ref: main
- name: Setup Rust environment
uses: ./.github/actions/setup
@@ -15,12 +15,14 @@
use crate::common::RustFSTestClusterEnvironment;
use aws_sdk_s3::Client;
use aws_sdk_s3::error::SdkError;
use aws_sdk_s3::types::{CorsConfiguration, CorsRule};
use bytes::Bytes;
use std::sync::Arc;
use tokio::sync::Barrier;
use tracing::{info, warn};
const BUCKET: &str = "conditional-put-race-bucket";
const BUCKET_METADATA_RELOAD_BUCKET: &str = "bucket-metadata-reload-barrier";
async fn cleanup_object(client: &Client, key: &str) {
if let Err(e) = client.delete_object().bucket(BUCKET).key(key).send().await {
@@ -28,6 +30,16 @@ async fn cleanup_object(client: &Client, key: &str) {
}
}
async fn assert_bucket_cors_missing(client: &Client) {
let result = client.get_bucket_cors().bucket(BUCKET_METADATA_RELOAD_BUCKET).send().await;
match result {
Err(SdkError::ServiceError(error)) => {
assert_eq!(error.err().meta().code(), Some("NoSuchCORSConfiguration"));
}
result => panic!("expected the peer to report a missing CORS configuration: {result:?}"),
}
}
async fn conditional_put(
client: &Client,
key: &str,
@@ -236,3 +248,48 @@ async fn test_conditional_put_basic_cluster() -> Result<(), Box<dyn std::error::
cleanup_object(&client, test_key).await;
Ok(())
}
#[tokio::test]
async fn test_bucket_cors_write_is_visible_on_peer_before_response() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
crate::common::init_logging();
let mut cluster = RustFSTestClusterEnvironment::new(2).await?;
cluster.start().await?;
cluster.create_test_bucket(BUCKET_METADATA_RELOAD_BUCKET).await?;
let writer = cluster.create_s3_client(0)?;
let reader = cluster.create_s3_client(1)?;
assert_bucket_cors_missing(&reader).await;
let rule = CorsRule::builder()
.allowed_methods("GET")
.allowed_origins("https://example.com")
.build()?;
let configuration = CorsConfiguration::builder().cors_rules(rule).build()?;
writer
.put_bucket_cors()
.bucket(BUCKET_METADATA_RELOAD_BUCKET)
.cors_configuration(configuration)
.send()
.await?;
let response = reader.get_bucket_cors().bucket(BUCKET_METADATA_RELOAD_BUCKET).send().await?;
let rules = response.cors_rules();
assert_eq!(
rules.len(),
1,
"peer should observe the committed CORS rule before the write response returns"
);
assert_eq!(rules[0].allowed_methods(), ["GET"]);
assert_eq!(rules[0].allowed_origins(), ["https://example.com"]);
writer
.delete_bucket_cors()
.bucket(BUCKET_METADATA_RELOAD_BUCKET)
.send()
.await?;
assert_bucket_cors_missing(&reader).await;
writer.delete_bucket().bucket(BUCKET_METADATA_RELOAD_BUCKET).send().await?;
Ok(())
}
@@ -85,6 +85,7 @@ const PEER_REST_RECOVERY_MAX_ATTEMPTS: u32 = 60;
const PEER_REST_RECOVERY_MAX_BACKOFF: Duration = Duration::from_secs(30);
const SCANNER_ACTIVITY_MAX_MESSAGE_SIZE: usize = 1024;
const REPLICATION_STATS_MAX_MESSAGE_SIZE: usize = 8 * 1024 * 1024;
const BUCKET_METADATA_RELOAD_TIMEOUT: Duration = Duration::from_secs(5);
/// Error for a peer that reported `success = false` without an `error_info` payload.
///
@@ -1328,27 +1329,38 @@ impl PeerRestClient {
}
pub async fn load_bucket_metadata(&self, bucket: &str, scanner_maintenance_change: bool) -> Result<()> {
self.finalize_result(
async {
let mut client = self.get_client().await?;
let mut request = Request::new(LoadBucketMetadataRequest {
bucket: bucket.to_string(),
scanner_maintenance_change,
});
set_tonic_mutation_body_digest(&mut request)?;
let response = client.load_bucket_metadata(request).await?.into_inner();
if !response.success {
if let Some(msg) = response.error_info {
return Err(Error::other(msg));
}
return Err(peer_failure_without_details("load_bucket_metadata", Some(bucket)));
}
Ok(())
let result = tokio::time::timeout(BUCKET_METADATA_RELOAD_TIMEOUT, async {
let result = self.load_bucket_metadata_once(bucket, scanner_maintenance_change).await;
if let Err(err) = &result
&& Self::is_network_like_error(err)
{
self.prepare_retry().await;
return self.load_bucket_metadata_once(bucket, scanner_maintenance_change).await;
}
.await,
)
result
})
.await
.unwrap_or_else(|_| Err(Error::other(format!("load_bucket_metadata({bucket}) timed out"))));
self.finalize_result(result).await
}
async fn load_bucket_metadata_once(&self, bucket: &str, scanner_maintenance_change: bool) -> Result<()> {
let mut client = self.get_client().await?;
let mut request = Request::new(LoadBucketMetadataRequest {
bucket: bucket.to_string(),
scanner_maintenance_change,
});
set_tonic_mutation_body_digest(&mut request)?;
request.set_timeout(BUCKET_METADATA_RELOAD_TIMEOUT);
let response = client.load_bucket_metadata(request).await?.into_inner();
if !response.success {
if let Some(msg) = response.error_info {
return Err(Error::other(msg));
}
return Err(peer_failure_without_details("load_bucket_metadata", Some(bucket)));
}
Ok(())
}
pub async fn delete_bucket_metadata(&self, bucket: &str) -> Result<()> {
+25 -86
View File
@@ -16,14 +16,14 @@
//!
//! `scripts/test/vault_ha_kms_live.sh` owns the official Vault containers and
//! kills the active node while this test continuously decrypts through a
//! surviving standby. KV2 and Transit must recover after the bounded circuit
//! interval, use a bounded number of attempts, and leave the circuit and
//! in-flight gauges at zero after a new leader is elected.
//! surviving standby. KV2 and Transit requests must remain successful, use a
//! bounded number of attempts, and leave the circuit and in-flight gauges at
//! zero after a new leader is elected.
use std::collections::HashMap;
use std::path::{Path, PathBuf};
use std::sync::Arc;
use std::sync::atomic::{AtomicBool, AtomicU64, Ordering};
use std::sync::{Arc, Mutex};
use std::time::Duration;
use metrics_util::MetricKind;
@@ -43,11 +43,6 @@ const OPERATION_ATTEMPTS: &str = "rustfs_kms_backend_operation_attempts";
const IN_FLIGHT: &str = "rustfs_kms_backend_in_flight";
const CIRCUIT_OPEN: &str = "rustfs_kms_backend_circuit_open";
const MAX_ATTEMPTS: u32 = 10;
const ATTEMPT_TIMEOUT: Duration = Duration::from_secs(2);
const HEALTHY_PROGRESS_TIMEOUT: Duration = Duration::from_secs(20);
// The circuit remains open for 30s after five failed attempts.
const POST_FAILOVER_PROGRESS_TIMEOUT: Duration = Duration::from_secs(35);
const FAILOVER_ERROR_POLL_INTERVAL: Duration = Duration::from_millis(100);
type MetricEntry = (
metrics_util::CompositeKey,
@@ -69,7 +64,7 @@ fn config(backend: KmsBackend, backend_config: BackendConfig) -> KmsConfig {
backend,
backend_config,
allow_insecure_dev_defaults: true,
timeout: ATTEMPT_TIMEOUT,
timeout: Duration::from_secs(2),
retry_attempts: MAX_ATTEMPTS,
enable_cache: false,
..KmsConfig::default()
@@ -169,31 +164,14 @@ fn retryable_failures(snapshot: &[MetricEntry], operation: &str) -> u64 {
.sum()
}
async fn wait_for_count(
counter: &AtomicU64,
failure: &Mutex<Option<String>>,
minimum: u64,
description: &str,
timeout: Duration,
) {
tokio::time::timeout(timeout, async {
async fn wait_for_count(counter: &AtomicU64, minimum: u64, description: &str) {
tokio::time::timeout(Duration::from_secs(20), async {
while counter.load(Ordering::SeqCst) < minimum {
if let Some(error) = failure.lock().expect("decrypt failure lock poisoned").as_ref() {
panic!(
"{description} worker failed after {} successful decrypts: {error}",
counter.load(Ordering::SeqCst)
);
}
tokio::time::sleep(Duration::from_millis(25)).await;
}
})
.await
.unwrap_or_else(|_| {
panic!(
"timed out after {timeout:?} waiting for {description}: completed {}, expected {minimum}",
counter.load(Ordering::SeqCst)
)
});
.unwrap_or_else(|_| panic!("timed out waiting for {description}"));
}
async fn wait_for_file(path: &Path, description: &str) {
@@ -211,8 +189,7 @@ async fn decrypt_loop<B: KmsBackendTrait + Send + Sync + 'static>(
request: DecryptRequest,
expected: Vec<u8>,
completed: Arc<AtomicU64>,
allow_failover_errors: Arc<AtomicBool>,
failure: Arc<Mutex<Option<String>>>,
failed: Arc<AtomicBool>,
stop: CancellationToken,
) {
while !stop.is_cancelled() {
@@ -220,18 +197,8 @@ async fn decrypt_loop<B: KmsBackendTrait + Send + Sync + 'static>(
Ok(response) if response.plaintext == expected => {
completed.fetch_add(1, Ordering::SeqCst);
}
Ok(_) => {
*failure.lock().expect("decrypt failure lock poisoned") =
Some("decrypt returned unexpected plaintext".to_string());
return;
}
Err(rustfs_kms::KmsError::BackendError { .. } | rustfs_kms::KmsError::OperationTimedOut { .. })
if allow_failover_errors.load(Ordering::SeqCst) =>
{
tokio::time::sleep(FAILOVER_ERROR_POLL_INTERVAL).await;
}
Err(error) => {
*failure.lock().expect("decrypt failure lock poisoned") = Some(error.to_string());
Ok(_) | Err(_) => {
failed.store(true, Ordering::SeqCst);
return;
}
}
@@ -329,9 +296,7 @@ async fn exercise_failover(snapshotter: &Snapshotter) {
);
let stop = CancellationToken::new();
let allow_failover_errors = Arc::new(AtomicBool::new(false));
let kv2_failure = Arc::new(Mutex::new(None));
let transit_failure = Arc::new(Mutex::new(None));
let failed = Arc::new(AtomicBool::new(false));
let kv2_completed = Arc::new(AtomicU64::new(0));
let transit_completed = Arc::new(AtomicU64::new(0));
let kv2_worker = tokio::spawn(decrypt_loop(
@@ -339,8 +304,7 @@ async fn exercise_failover(snapshotter: &Snapshotter) {
kv2_request,
kv2_data_key.plaintext_key,
Arc::clone(&kv2_completed),
Arc::clone(&allow_failover_errors),
Arc::clone(&kv2_failure),
Arc::clone(&failed),
stop.clone(),
));
let transit_worker = tokio::spawn(decrypt_loop(
@@ -348,21 +312,12 @@ async fn exercise_failover(snapshotter: &Snapshotter) {
transit_request,
transit_data_key.plaintext_key,
Arc::clone(&transit_completed),
Arc::clone(&allow_failover_errors),
Arc::clone(&transit_failure),
Arc::clone(&failed),
stop.clone(),
));
wait_for_count(&kv2_completed, &kv2_failure, 2, "two healthy KV2 decrypts", HEALTHY_PROGRESS_TIMEOUT).await;
wait_for_count(
&transit_completed,
&transit_failure,
2,
"two healthy Transit decrypts",
HEALTHY_PROGRESS_TIMEOUT,
)
.await;
allow_failover_errors.store(true, Ordering::SeqCst);
wait_for_count(&kv2_completed, 2, "two healthy KV2 decrypts").await;
wait_for_count(&transit_completed, 2, "two healthy Transit decrypts").await;
std::fs::write(&marker, b"ready").expect("publish failover readiness marker");
wait_for_file(&elected, "the replacement Vault leader").await;
@@ -371,39 +326,18 @@ async fn exercise_failover(snapshotter: &Snapshotter) {
let kv2_after_election = kv2_completed.load(Ordering::SeqCst) + 2;
let transit_after_election = transit_completed.load(Ordering::SeqCst) + 2;
wait_for_count(
&kv2_completed,
&kv2_failure,
kv2_after_election,
"post-failover KV2 decrypts",
POST_FAILOVER_PROGRESS_TIMEOUT,
)
.await;
wait_for_count(
&transit_completed,
&transit_failure,
transit_after_election,
"post-failover Transit decrypts",
POST_FAILOVER_PROGRESS_TIMEOUT,
)
.await;
wait_for_count(&kv2_completed, kv2_after_election, "post-failover KV2 decrypts").await;
wait_for_count(&transit_completed, transit_after_election, "post-failover Transit decrypts").await;
stop.cancel();
kv2_worker.await.expect("KV2 decrypt worker must join");
transit_worker.await.expect("Transit decrypt worker must join");
assert!(
kv2_failure.lock().expect("KV2 failure lock poisoned").is_none(),
"no KV2 decrypt may fail or return different plaintext"
);
assert!(
transit_failure.lock().expect("Transit failure lock poisoned").is_none(),
"no Transit decrypt may fail or return different plaintext"
);
assert!(!failed.load(Ordering::SeqCst), "no decrypt may fail or return different plaintext");
}
#[test]
#[ignore = "requires a real three-node Vault Raft cluster; run scripts/test/vault_ha_kms_live.sh"]
fn vault_raft_leader_failure_recovers_kv2_and_transit_decrypts() {
fn vault_raft_leader_failure_preserves_kv2_and_transit_decrypts() {
let recorder = DebuggingRecorder::new();
let snapshotter = recorder.snapshotter();
metrics::with_local_recorder(&recorder, || {
@@ -415,6 +349,11 @@ fn vault_raft_leader_failure_recovers_kv2_and_transit_decrypts() {
});
let snapshot = snapshotter.snapshot().into_vec();
assert_eq!(
counter_value(&snapshot, OPERATIONS_TOTAL, &[("outcome", "circuit_open")]),
0,
"a bounded leader election must not open the circuit"
);
assert_eq!(
counter_value(&snapshot, OPERATIONS_TOTAL, &[("outcome", "budget_exhausted")]),
0,
+4 -11
View File
@@ -158,11 +158,10 @@ added by backlog#1153 infra-4.
## Coverage
Workspace line coverage is measured weekly. Pull requests that touch iam, kms,
policy, or crypto also run a non-required, report-only comparison against
`.config/coverage-baselines.toml`. During calibration, a regression is recorded
in the job summary without failing the job; missing or malformed coverage
evidence still fails closed (backlog#1153 infra-6).
Line coverage is measured **weekly, not per-PR**, and is non-blocking: it
exists for visibility and trend, never as a required check. Per-crate ratchets
for the security-critical crates (iam / kms / policy / crypto) build on this
baseline later (backlog#1153 infra-6, report-only first).
- **CI**: `.github/workflows/coverage.yml` runs every Sunday and on manual
dispatch: `cargo llvm-cov nextest --workspace --exclude e2e_test` under the
@@ -175,12 +174,6 @@ evidence still fails closed (backlog#1153 infra-6).
plus the full suite). It prints the same per-crate table via
`scripts/coverage_per_crate.py` and writes `target/llvm-cov/lcov.info` and
`coverage.json`.
- **Security-critical ratchet**: relevant pull requests compare iam / kms /
policy / crypto line coverage with the versioned baseline. Drops greater than
the configured one-percentage-point calibration threshold are marked
`REGRESSION (report-only)`. The weekly summary runs the same comparison so
calibration continues even when no relevant pull request is open. Baseline
changes require a linked coverage run and a reviewed explanation.
- **Trend comparison**: each run's job summary is the weekly per-crate
snapshot — open two runs from the Actions history (workflow "coverage") and
compare their tables. For line-level diffs, download the two runs'
+22 -18
View File
@@ -513,13 +513,15 @@ fn sr_bucket_meta_item(bucket: String, item_type: &str) -> SRBucketMeta {
}
}
fn notify_bucket_metadata_reload(
async fn notify_bucket_metadata_reload(
bucket: String,
operation: &'static str,
request_context: Option<request_context::RequestContext>,
scanner_maintenance_change: bool,
) {
record_local_scanner_maintenance_reload(&bucket, scanner_maintenance_change);
// Keep reload detached across request cancellation, but wait before a healthy peer can serve the previous config.
let (completed_tx, completed_rx) = tokio::sync::oneshot::channel();
spawn_background_with_context(request_context, async move {
if let Some(notification_sys) = current_notification_system() {
let result = if scanner_maintenance_change {
@@ -531,7 +533,9 @@ fn notify_bucket_metadata_reload(
warn!(bucket = %bucket, error = %err, "failed to notify peers after {operation}");
}
}
let _ = completed_tx.send(());
});
let _ = completed_rx.await;
}
fn record_local_scanner_maintenance_reload(bucket: &str, scanner_maintenance_change: bool) {
@@ -1476,7 +1480,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "delete bucket encryption", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "delete bucket encryption", request_context, false).await;
let item = sr_bucket_meta_item(bucket.clone(), "sse-config");
if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1508,7 +1512,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "delete bucket cors", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "delete bucket cors", request_context, false).await;
let item = sr_bucket_meta_item(bucket.clone(), "cors-config");
if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1540,7 +1544,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "delete bucket lifecycle", request_context, true);
notify_bucket_metadata_reload(bucket.clone(), "delete bucket lifecycle", request_context, true).await;
let item = sr_bucket_meta_item(bucket.clone(), "lc-config");
if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1572,7 +1576,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "delete bucket policy", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "delete bucket policy", request_context, false).await;
let item = sr_bucket_meta_item(bucket.clone(), "policy");
if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1630,7 +1634,7 @@ impl DefaultBucketUsecase {
}
drop(targets_guard);
notify_bucket_metadata_reload(bucket.clone(), "delete bucket replication", request_context, true);
notify_bucket_metadata_reload(bucket.clone(), "delete bucket replication", request_context, true).await;
let item = sr_bucket_meta_item(bucket.clone(), "replication-config");
if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1655,7 +1659,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "delete bucket tagging", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "delete bucket tagging", request_context, false).await;
let item = sr_bucket_meta_item(bucket.clone(), "tags");
if let Err(err) = site_replication_bucket_meta_hook(item).await {
@@ -1688,7 +1692,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "delete public access block", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "delete public access block", request_context, false).await;
Ok(S3Response::with_status(DeletePublicAccessBlockOutput::default(), StatusCode::NO_CONTENT))
}
@@ -2143,7 +2147,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "put bucket encryption", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "put bucket encryption", request_context, false).await;
let mut item = sr_bucket_meta_item(bucket.clone(), "sse-config");
item.sse_config = Some(
@@ -2222,7 +2226,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "put bucket lifecycle", request_context, true);
notify_bucket_metadata_reload(bucket.clone(), "put bucket lifecycle", request_context, true).await;
let mut item = sr_bucket_meta_item(bucket.clone(), "lc-config");
item.expiry_lc_config =
@@ -2307,7 +2311,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "put bucket notification", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "put bucket notification", request_context, false).await;
let region = resolve_notification_region(self.global_region(), request_region);
let notify = current_notify_interface_for_context(self.context.as_deref());
@@ -2412,7 +2416,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "put bucket policy", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "put bucket policy", request_context, false).await;
let mut item = sr_bucket_meta_item(bucket.clone(), "policy");
item.policy = Some(serde_json::from_str(&policy).map_err(|e| s3_error!(InvalidArgument, "parse policy failed {:?}", e))?);
@@ -2447,7 +2451,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "put bucket cors", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "put bucket cors", request_context, false).await;
let mut item = sr_bucket_meta_item(bucket.clone(), "cors-config");
item.cors =
@@ -2491,7 +2495,7 @@ impl DefaultBucketUsecase {
.map_err(ApiError::from)?;
drop(targets_guard);
notify_bucket_metadata_reload(bucket.clone(), "put bucket replication", request_context, true);
notify_bucket_metadata_reload(bucket.clone(), "put bucket replication", request_context, true).await;
let mut item = sr_bucket_meta_item(bucket.clone(), "replication-config");
item.replication_config = Some(
@@ -2531,7 +2535,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "put public access block", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "put public access block", request_context, false).await;
Ok(S3Response::new(PutPublicAccessBlockOutput::default()))
}
@@ -2560,7 +2564,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "put bucket tagging", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "put bucket tagging", request_context, false).await;
let mut item = sr_bucket_meta_item(bucket.clone(), "tags");
item.tags = Some(serialize_config(&tagging).and_then(|bytes| String::from_utf8(bytes).map_err(to_internal_error))?);
@@ -2593,7 +2597,7 @@ impl DefaultBucketUsecase {
.await
.map_err(ApiError::from)?;
notify_bucket_metadata_reload(bucket.clone(), "put bucket versioning", request_context, false);
notify_bucket_metadata_reload(bucket.clone(), "put bucket versioning", request_context, false).await;
let mut item = sr_bucket_meta_item(bucket.clone(), "version-config");
item.versioning = Some(
@@ -3044,7 +3048,7 @@ mod tests {
"{method} should identify the bucket metadata operation in reload logs"
);
let expected_reload = format!(
"notify_bucket_metadata_reload(bucket.clone(), \"{operation}\", request_context, {scanner_maintenance_change});"
"notify_bucket_metadata_reload(bucket.clone(), \"{operation}\", request_context, {scanner_maintenance_change}).await;"
);
assert!(
body.contains(&expected_reload),
-264
View File
@@ -1,264 +0,0 @@
#!/usr/bin/env python3
# Copyright 2024 RustFS Team
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Compare security-critical crate line coverage with the report-only baseline."""
import argparse
import json
import math
import os
import sys
import tempfile
import tomllib
from pathlib import Path
from coverage_per_crate import fmt_pct, load_coverage
SECURITY_CRATES = ("crates/iam", "crates/kms", "crates/policy", "crates/crypto")
def load_baselines(path: str) -> tuple[float, dict[str, tuple[int, int]]]:
with open(path, "rb") as fh:
config = tomllib.load(fh)
if config.get("phase") != "report-only":
raise ValueError("coverage baseline phase must be report-only")
raw_allowed_drop = config["allowed_drop_percentage_points"]
if isinstance(raw_allowed_drop, bool) or not isinstance(raw_allowed_drop, (int, float)):
raise ValueError("allowed_drop_percentage_points must be a number")
allowed_drop = float(raw_allowed_drop)
if not math.isfinite(allowed_drop) or allowed_drop < 0:
raise ValueError("allowed_drop_percentage_points must be finite and non-negative")
baselines: dict[str, tuple[int, int]] = {}
for crate, values in config["crates"].items():
covered = values["covered"]
count = values["count"]
if type(covered) is not int or type(count) is not int:
raise ValueError(f"invalid baseline for {crate}: covered and count must be integers")
if covered < 0 or count <= 0 or covered > count:
raise ValueError(f"invalid baseline for {crate}: {covered}/{count}")
baselines[crate] = (covered, count)
missing = [crate for crate in SECURITY_CRATES if crate not in baselines]
unexpected = sorted(set(baselines).difference(SECURITY_CRATES))
if missing or unexpected:
raise ValueError(f"coverage baseline crate set mismatch: missing={missing}, unexpected={unexpected}")
return allowed_drop, baselines
def compare(
current: dict[str, list[int]],
baselines: dict[str, tuple[int, int]],
allowed_drop: float,
) -> list[tuple[str, int, int, int, int, float, bool]]:
rows = []
for crate, (baseline_covered, baseline_count) in baselines.items():
if crate not in current:
raise ValueError(f"coverage report is missing {crate}")
covered, count = current[crate]
if type(covered) is not int or type(count) is not int:
raise ValueError(f"invalid coverage for {crate}: covered and count must be integers")
if covered < 0 or count <= 0 or covered > count:
raise ValueError(f"invalid coverage for {crate}: {covered}/{count}")
current_pct = 100.0 * covered / count
baseline_pct = 100.0 * baseline_covered / baseline_count
delta = current_pct - baseline_pct
rows.append((crate, covered, count, baseline_covered, baseline_count, delta, delta < -allowed_drop))
return rows
def print_report(rows: list[tuple[str, int, int, int, int, float, bool]], allowed_drop: float) -> None:
print("## Security-critical coverage ratchet (report-only)")
print()
print(f"Calibration threshold: a drop greater than {allowed_drop:.2f} percentage points is reported as a regression.")
print()
print("| Crate | Current | Baseline | Delta | Status |")
print("|---|---:|---:|---:|---|")
for crate, covered, count, baseline_covered, baseline_count, delta, regressed in rows:
status = "REGRESSION (report-only)" if regressed else "OK"
print(
f"| `{crate}` | {fmt_pct(covered, count)} ({covered}/{count}) "
f"| {fmt_pct(baseline_covered, baseline_count)} ({baseline_covered}/{baseline_count}) "
f"| {delta:+.2f} pp | {status} |"
)
print()
print("This calibration phase records regressions without failing the job; malformed or incomplete evidence still fails closed.")
def self_test() -> None:
with tempfile.TemporaryDirectory() as tmp:
root = Path(tmp)
coverage = root / "coverage.json"
baseline = root / "baseline.toml"
coverage_data = {
"data": [
{
"files": [
{
"filename": str(root / "crates/iam/src/lib.rs"),
"summary": {"lines": {"covered": 80, "count": 100}},
},
{
"filename": str(root / "crates/kms/src/lib.rs"),
"summary": {"lines": {"covered": 90, "count": 100}},
},
{
"filename": str(root / "crates/policy/src/lib.rs"),
"summary": {"lines": {"covered": 90, "count": 100}},
},
{
"filename": str(root / "crates/crypto/src/lib.rs"),
"summary": {"lines": {"covered": 90, "count": 100}},
},
],
"totals": {"lines": {"covered": 350, "count": 400}},
}
]
}
coverage.write_text(json.dumps(coverage_data), encoding="utf-8")
baseline_text = """phase = "report-only"
allowed_drop_percentage_points = 1.0
[crates."crates/iam"]
covered = 90
count = 100
[crates."crates/kms"]
covered = 85
count = 100
[crates."crates/policy"]
covered = 90
count = 100
[crates."crates/crypto"]
covered = 90
count = 100
"""
baseline.write_text(baseline_text, encoding="utf-8")
current, _ = load_coverage(str(coverage), str(root))
allowed_drop, baselines = load_baselines(str(baseline))
rows = compare(current, baselines, allowed_drop)
assert [row[-1] for row in rows] == [True, False, False, False]
try:
compare({"crates/iam": current["crates/iam"]}, baselines, allowed_drop)
except ValueError as error:
assert str(error) == "coverage report is missing crates/kms"
else:
raise AssertionError("missing crate must fail closed")
try:
compare({**current, "crates/iam": [101, 100]}, baselines, allowed_drop)
except ValueError as error:
assert str(error) == "invalid coverage for crates/iam: 101/100"
else:
raise AssertionError("invalid coverage must fail closed")
for invalid_threshold in ("true", '"1.0"', "nan", "inf", "-inf"):
baseline.write_text(
baseline_text.replace("allowed_drop_percentage_points = 1.0", f"allowed_drop_percentage_points = {invalid_threshold}"),
encoding="utf-8",
)
try:
load_baselines(str(baseline))
except ValueError:
pass
else:
raise AssertionError(f"non-finite threshold {invalid_threshold} must fail closed")
for field, invalid_values in (
("covered", ("true", '"90"', "90.0", "90.5")),
("count", ("true", '"100"', "100.0", "100.5")),
):
for invalid_value in invalid_values:
baseline.write_text(
baseline_text.replace(f"{field} = {90 if field == 'covered' else 100}", f"{field} = {invalid_value}", 1),
encoding="utf-8",
)
try:
load_baselines(str(baseline))
except ValueError:
pass
else:
raise AssertionError(f"non-integer baseline {field} {invalid_value} must fail closed")
for covered, count in (
(True, 100),
(80, True),
(80.0, 100),
(80, 100.0),
(float("nan"), 100),
(80, float("inf")),
):
try:
compare({**current, "crates/iam": [covered, count]}, baselines, allowed_drop)
except ValueError:
pass
else:
raise AssertionError(f"invalid aggregate coverage {covered}/{count} must fail closed")
lines = coverage_data["data"][0]["files"][0]["summary"]["lines"]
for field, invalid_values in (
("covered", (True, "80", 80.0, 80.5, float("nan"), float("inf"), float("-inf"))),
("count", (True, "100", 100.0, 100.5, float("nan"), float("inf"), float("-inf"))),
):
original = lines[field]
for invalid_value in invalid_values:
lines[field] = invalid_value
coverage.write_text(json.dumps(coverage_data), encoding="utf-8")
try:
load_coverage(str(coverage), str(root))
except ValueError:
pass
else:
raise AssertionError(f"invalid raw coverage {field} {invalid_value} must fail closed")
lines[field] = original
baseline.write_text(
baseline_text.replace(
'[crates."crates/crypto"]\ncovered = 90\ncount = 100\n',
"",
),
encoding="utf-8",
)
try:
load_baselines(str(baseline))
except ValueError:
pass
else:
raise AssertionError("missing security-crate baseline must fail closed")
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("coverage_json", nargs="?")
parser.add_argument("--baseline", default=".config/coverage-baselines.toml")
parser.add_argument("--repo-root", default=os.getcwd())
parser.add_argument("--self-test", action="store_true")
args = parser.parse_args()
if args.self_test:
self_test()
print("security coverage self-test passed")
return 0
if not args.coverage_json:
parser.error("coverage_json is required unless --self-test is used")
try:
current, _ = load_coverage(args.coverage_json, os.path.abspath(args.repo_root))
allowed_drop, baselines = load_baselines(args.baseline)
rows = compare(current, baselines, allowed_drop)
except (OSError, ValueError, KeyError, IndexError, json.JSONDecodeError, tomllib.TOMLDecodeError) as error:
print(f"error: {error}", file=sys.stderr)
return 1
print_report(rows, allowed_drop)
return 0
if __name__ == "__main__":
sys.exit(main())
+14 -27
View File
@@ -47,31 +47,6 @@ def fmt_pct(covered: int, count: int) -> str:
return f"{100.0 * covered / count:.2f}%" if count else ""
def _line_counts(lines: dict[str, int], source: str) -> tuple[int, int]:
covered = lines["covered"]
count = lines["count"]
if type(covered) is not int or type(count) is not int or covered < 0 or count < 0 or covered > count:
raise ValueError(f"invalid line coverage for {source}: {covered}/{count}")
return covered, count
def load_coverage(path: str, root: str) -> tuple[dict[str, list[int]], dict[str, int]]:
with open(path, encoding="utf-8") as fh:
export = json.load(fh)
data = export["data"][0]
files = data["files"]
total_covered, total_count = _line_counts(data["totals"]["lines"], "totals")
crates: dict[str, list[int]] = {}
for f in files:
covered, count = _line_counts(f["summary"]["lines"], f["filename"])
acc = crates.setdefault(crate_label(f["filename"], root), [0, 0])
acc[0] += covered
acc[1] += count
return crates, {"covered": total_covered, "count": total_count}
def main() -> int:
if len(sys.argv) < 2 or len(sys.argv) > 3:
print(__doc__.strip(), file=sys.stderr)
@@ -79,12 +54,24 @@ def main() -> int:
path = sys.argv[1]
root = os.path.abspath(sys.argv[2] if len(sys.argv) == 3 else os.getcwd())
with open(path, encoding="utf-8") as fh:
export = json.load(fh)
try:
crates, totals = load_coverage(path, root)
except (KeyError, IndexError, ValueError) as exc:
data = export["data"][0]
files = data["files"]
totals = data["totals"]["lines"]
except (KeyError, IndexError) as exc:
print(f"error: unexpected llvm-cov JSON shape ({exc})", file=sys.stderr)
return 1
crates: dict[str, list[int]] = {}
for f in files:
lines = f["summary"]["lines"]
acc = crates.setdefault(crate_label(f["filename"], root), [0, 0])
acc[0] += lines["covered"]
acc[1] += lines["count"]
rows = sorted(
crates.items(),
key=lambda kv: (100.0 * kv[1][0] / kv[1][1]) if kv[1][1] else 101.0,
+1 -1
View File
@@ -241,7 +241,7 @@ env \
RUSTFS_TEST_VAULT_FAILOVER_MARKER="$MARKER" \
RUSTFS_TEST_VAULT_OLD_LEADER="$OLD_LEADER" \
cargo test -p rustfs-kms --test vault_ha_failover_live \
vault_raft_leader_failure_recovers_kv2_and_transit_decrypts -- \
vault_raft_leader_failure_preserves_kv2_and_transit_decrypts -- \
--ignored --nocapture --test-threads=1 &
TEST_PID=$!