gsadmin/rustfs
1
0
Fork 0
mirror of https://github.com/rustfs/rustfs.git synced 2026-09-07 20:46:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(scanner): recover legacy usage floor from backup (#6964)

Browse Source 
* fix(scanner): recover legacy usage floor from backup

Allow scanner usage-floor startup and leadership fencing to use a valid legacy backup when the legacy primary read fails with a corruption-shaped error.

Keep v2 primary read failures, stale metadata, transient I/O, and missing or invalid backups fail-closed.

Co-Authored-By: heihutu <heihutu@gmail.com>

* fix(scanner): cover legacy backup fencing gaps (#6966)

* fix(scanner): recover legacy usage from valid backup

* fix(scanner): recover legacy usage floor from backup

Allow scanner usage-floor startup and leadership fencing to use a valid legacy backup when the legacy primary read fails with a corruption-shaped error.

Keep v2 primary read failures, stale metadata, transient I/O, and missing or invalid backups fail-closed.

Co-Authored-By: heihutu <heihutu@gmail.com>

* fix(scanner): cover legacy backup fencing gaps

---------

Co-authored-by: houseme <housemecn@gmail.com>
Co-authored-by: heihutu <heihutu@gmail.com>

---------

Co-authored-by: heihutu <heihutu@gmail.com>
Co-authored-by: Henry Guo <marshawcoco@gmail.com>
This commit is contained in:
houseme
2026-09-01 00:01:38 +08:00
committed by GitHub
parent e3ca1ca54c
commit e44007012b
5 changed files with 421 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crates/scanner/src/scanner/usage_store.rs
+27 -2
View File
@@ -13,6 +13,7 @@
// limitations under the License.
/// Data-usage snapshot persistence: CAS store pipeline, epoch baselines, and observed-snapshot cleanup.
use super::*;
use crate::data_usage_define::usage_floor_primary_read_error_allows_backup;
use crate::storage_api::owner::ScannerPublicationCommitState;
use std::collections::HashMap;
use std::sync::atomic::AtomicBool;
@@ -53,6 +54,30 @@ pub(super) struct DataUsagePersistBaseline {
pub(super) revision: DataUsageCacheRevision,
}
async fn read_usage_persist_candidate(
storeapi: Arc<impl ScannerObjectIO>,
path: &str,
) -> Result<(Option<Vec<u8>>, DataUsageCacheRevision), EcstoreError> {
let primary = read_config_with_revision(storeapi.clone(), path).await;
if path != LEGACY_DATA_USAGE_OBJ_NAME_PATH.as_str() {
return primary;
}
let Err(primary_error) = &primary else {
return primary;
};
if !usage_floor_primary_read_error_allows_backup(primary_error) {
return primary;
}
let backup_path = format!("{path}.bkp");
let backup = read_config_with_revision(storeapi, &backup_path).await?;
if backup.0.as_deref().is_some_and(|data| {
serde_json::from_slice::<DataUsageInfo>(data).is_ok_and(|usage| data_usage_info_has_persisted_baseline_identity(&usage))
}) {
return Ok(backup);
}
primary
}
/// Read the bytes used as the baseline for a usage publication while keeping
/// the v2 primary revision as the CAS fence. During an interrupted upgrade the
/// primary can be valid JSON without a baseline identity; in that case a
@@ -68,7 +93,7 @@ pub(super) async fn read_data_usage_persist_baseline(
LEGACY_DATA_USAGE_OBJ_NAME_PATH.as_str().to_string(),
format!("{}.bkp", LEGACY_DATA_USAGE_OBJ_NAME_PATH.as_str()),
] {
let (candidate, _) = read_config_with_revision(storeapi.clone(), &path).await?;
let (candidate, _) = read_usage_persist_candidate(storeapi.clone(), &path).await?;
let Some(candidate) = candidate else {
continue;
};
@@ -107,7 +132,7 @@ pub(super) async fn read_data_usage_persist_baseline(
LEGACY_DATA_USAGE_OBJ_NAME_PATH.as_str().to_string(),
format!("{}.bkp", LEGACY_DATA_USAGE_OBJ_NAME_PATH.as_str()),
] {
let (candidate, _) = read_config_with_revision(storeapi.clone(), &path).await?;
let (candidate, _) = read_usage_persist_candidate(storeapi.clone(), &path).await?;
let Some(candidate) = candidate else {
continue;
};