mirror of
https://github.com/rustfs/rustfs.git
synced 2026-09-01 09:48:20 +00:00
fix(ecstore): classify system metadata failures (#6862)
fix(ecstore): classify system metadata volume failures Preserve retryable quorum errors when system metadata reads or writes encounter missing volumes, and cover the create-bucket data-usage path with regressions. Co-authored-by: heihutu <heihutu@gmail.com>
This commit is contained in:
@@ -800,11 +800,22 @@ where
|
||||
if log_error {
|
||||
error!("save_config_with_opts: err: {:?}, file: {}", err, file);
|
||||
}
|
||||
Err(err)
|
||||
Err(map_system_metadata_write_error(err, file))
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// A system metadata volume outage must remain retryable instead of being
|
||||
/// exposed as the user-facing bucket-not-found response.
|
||||
pub(crate) fn map_system_metadata_write_error(err: Error, file: &str) -> Error {
|
||||
match err {
|
||||
Error::BucketNotFound(_) | Error::VolumeNotFound => {
|
||||
Error::InsufficientWriteQuorum(RUSTFS_META_BUCKET.to_string(), file.to_string())
|
||||
}
|
||||
other => other,
|
||||
}
|
||||
}
|
||||
|
||||
fn new_server_config() -> Config {
|
||||
Config::new()
|
||||
}
|
||||
@@ -2796,14 +2807,14 @@ mod tests {
|
||||
use super::{
|
||||
SERVER_CONFIG_LOCK, ServerConfigSnapshot, apply_dynamic_config_for_sub_sys_with, build_scalar_config_object,
|
||||
config_task_join_error, configs_semantically_equal, decode_server_config_blob, encode_server_config_blob,
|
||||
heal_config_descriptor, is_standard_object_server_config, lookup_configs, new_and_save_server_config, read_config,
|
||||
read_config_no_lock_preserve_empty_with_metadata, read_config_preserve_empty, read_config_with_metadata,
|
||||
read_config_without_migrate, read_server_config_snapshot, save_server_config, save_server_config_snapshot,
|
||||
save_server_config_snapshot_with_generation, server_config_transaction_lock_path, should_warn_ignored_scalar_section,
|
||||
storage_class_kvs_mut,
|
||||
heal_config_descriptor, is_standard_object_server_config, lookup_configs, map_system_metadata_write_error,
|
||||
new_and_save_server_config, read_config, read_config_no_lock_preserve_empty_with_metadata, read_config_preserve_empty,
|
||||
read_config_with_metadata, read_config_without_migrate, read_server_config_snapshot, save_config_with_opts_inner,
|
||||
save_server_config, save_server_config_snapshot, save_server_config_snapshot_with_generation,
|
||||
server_config_transaction_lock_path, should_warn_ignored_scalar_section, storage_class_kvs_mut,
|
||||
};
|
||||
use crate::config::{audit, heal, notify, oidc, scanner};
|
||||
use crate::disk::endpoint::Endpoint;
|
||||
use crate::disk::{RUSTFS_META_BUCKET, endpoint::Endpoint};
|
||||
use crate::error::{Error, Result};
|
||||
use crate::layout::endpoints::SetupType;
|
||||
use crate::object_api::{GetObjectReader, ObjectInfo, ObjectOptions, PutObjReader};
|
||||
@@ -2835,6 +2846,72 @@ mod tests {
|
||||
assert!(rendered.contains("panicked"));
|
||||
assert!(!rendered.contains("do-not-expose-payload"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn system_metadata_volume_failures_map_to_retryable_write_errors() {
|
||||
for error in [Error::VolumeNotFound, Error::BucketNotFound(RUSTFS_META_BUCKET.to_string())] {
|
||||
assert_eq!(
|
||||
map_system_metadata_write_error(error, "buckets/example/.metadata.bin"),
|
||||
Error::InsufficientWriteQuorum(RUSTFS_META_BUCKET.to_string(), "buckets/example/.metadata.bin".to_string())
|
||||
);
|
||||
}
|
||||
|
||||
let other = Error::other("metadata encoding failed");
|
||||
assert_eq!(map_system_metadata_write_error(other.clone(), "buckets/example/.metadata.bin"), other);
|
||||
}
|
||||
|
||||
#[derive(Debug, Default)]
|
||||
struct MetadataWriteStore {
|
||||
error: Option<Error>,
|
||||
}
|
||||
|
||||
#[async_trait::async_trait]
|
||||
impl crate::storage_api_contracts::object::ObjectIO for MetadataWriteStore {
|
||||
type Error = Error;
|
||||
type RangeSpec = HTTPRangeSpec;
|
||||
type HeaderMap = HeaderMap;
|
||||
type ObjectOptions = ObjectOptions;
|
||||
type ObjectInfo = ObjectInfo;
|
||||
type GetObjectReader = GetObjectReader;
|
||||
type PutObjectReader = PutObjReader;
|
||||
|
||||
async fn get_object_reader(
|
||||
&self,
|
||||
_bucket: &str,
|
||||
_object: &str,
|
||||
_range: Option<Self::RangeSpec>,
|
||||
_headers: Self::HeaderMap,
|
||||
_opts: &Self::ObjectOptions,
|
||||
) -> core::result::Result<Self::GetObjectReader, Self::Error> {
|
||||
Err(Error::FileNotFound)
|
||||
}
|
||||
|
||||
async fn put_object(
|
||||
&self,
|
||||
_bucket: &str,
|
||||
_object: &str,
|
||||
_data: &mut Self::PutObjectReader,
|
||||
_opts: &Self::ObjectOptions,
|
||||
) -> core::result::Result<Self::ObjectInfo, Self::Error> {
|
||||
Err(self.error.clone().expect("test store error should be configured"))
|
||||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn save_config_preserves_retryable_system_volume_errors() {
|
||||
let store = Arc::new(MetadataWriteStore {
|
||||
error: Some(Error::BucketNotFound(RUSTFS_META_BUCKET.to_string())),
|
||||
});
|
||||
let error =
|
||||
save_config_with_opts_inner(store, "buckets/example/.metadata.bin", Vec::new(), &ObjectOptions::default(), false)
|
||||
.await
|
||||
.expect_err("missing metadata volume must fail");
|
||||
|
||||
assert_eq!(
|
||||
error,
|
||||
Error::InsufficientWriteQuorum(RUSTFS_META_BUCKET.to_string(), "buckets/example/.metadata.bin".to_string())
|
||||
);
|
||||
}
|
||||
use rustfs_lock::client::LockClient;
|
||||
use rustfs_lock::client::local::LocalClient;
|
||||
use rustfs_lock::{LockError, LockInfo, LockResponse, LockStats};
|
||||
|
||||
@@ -422,9 +422,7 @@ async fn save_data_usage_in_backend(
|
||||
if publication_epoch != expected_publication_epoch {
|
||||
return Err(Error::other("data usage publication epoch changed before save"));
|
||||
}
|
||||
crate::config::com::save_config(store.clone(), &DATA_USAGE_OBJ_NAME_PATH, data)
|
||||
.await
|
||||
.map_err(Error::other)?;
|
||||
crate::config::com::save_config(store.clone(), &DATA_USAGE_OBJ_NAME_PATH, data).await?;
|
||||
drop(publication_guard);
|
||||
|
||||
cleanup_observed_data_usage_after_authoritative_save_with_publication(store.as_ref(), &data_usage_info, Some(store.as_ref()))
|
||||
@@ -641,7 +639,7 @@ where
|
||||
{
|
||||
Ok(reader) => reader,
|
||||
Err(Error::FileNotFound | Error::ObjectNotFound(_, _) | Error::ConfigNotFound) => return Ok(None),
|
||||
Err(err) => return Err(err),
|
||||
Err(err) => return Err(map_data_usage_metadata_read_error(err, object)),
|
||||
};
|
||||
let revision = reader
|
||||
.object_info
|
||||
@@ -656,6 +654,18 @@ where
|
||||
Ok(Some((data_usage_info, revision)))
|
||||
}
|
||||
|
||||
/// A missing usage object is harmless during bucket creation, but a missing
|
||||
/// system metadata volume is a storage outage. Keep the latter retryable and
|
||||
/// distinguishable from the user bucket not existing.
|
||||
fn map_data_usage_metadata_read_error(err: Error, object: &str) -> Error {
|
||||
match err {
|
||||
Error::BucketNotFound(_) | Error::VolumeNotFound => {
|
||||
Error::InsufficientReadQuorum(RUSTFS_META_BUCKET.to_string(), object.to_string())
|
||||
}
|
||||
other => other,
|
||||
}
|
||||
}
|
||||
|
||||
fn data_usage_contains_bucket(data_usage_info: &DataUsageInfo, bucket: &str) -> bool {
|
||||
data_usage_info.buckets_usage.contains_key(bucket) || data_usage_info.bucket_sizes.contains_key(bucket)
|
||||
}
|
||||
@@ -912,7 +922,7 @@ where
|
||||
)
|
||||
.await;
|
||||
drop(publication_guard);
|
||||
match save_result {
|
||||
match save_result.map_err(|err| crate::config::com::map_system_metadata_write_error(err, object)) {
|
||||
Ok(_) => return Ok(()),
|
||||
Err(err) => {
|
||||
if let Some((observed, observed_revision)) = load_data_usage_for_bucket_removal(store, object).await? {
|
||||
@@ -2759,6 +2769,7 @@ mod tests {
|
||||
struct UsageCacheReadStore {
|
||||
transient_failures: Mutex<usize>,
|
||||
reads: Mutex<Vec<String>>,
|
||||
terminal_error: Mutex<Option<Error>>,
|
||||
}
|
||||
|
||||
impl UsageCacheReadStore {
|
||||
@@ -2766,6 +2777,15 @@ mod tests {
|
||||
Self {
|
||||
transient_failures: Mutex::new(n),
|
||||
reads: Mutex::new(Vec::new()),
|
||||
terminal_error: Mutex::new(None),
|
||||
}
|
||||
}
|
||||
|
||||
fn with_terminal_error(error: Error) -> Self {
|
||||
Self {
|
||||
transient_failures: Mutex::new(0),
|
||||
reads: Mutex::new(Vec::new()),
|
||||
terminal_error: Mutex::new(Some(error)),
|
||||
}
|
||||
}
|
||||
|
||||
@@ -2793,6 +2813,9 @@ mod tests {
|
||||
_opts: &Self::ObjectOptions,
|
||||
) -> Result<Self::GetObjectReader, Self::Error> {
|
||||
self.reads.lock().await.push(object.to_string());
|
||||
if let Some(error) = self.terminal_error.lock().await.clone() {
|
||||
return Err(error);
|
||||
}
|
||||
let mut remaining = self.transient_failures.lock().await;
|
||||
if *remaining > 0 {
|
||||
*remaining -= 1;
|
||||
@@ -2857,6 +2880,22 @@ mod tests {
|
||||
assert!(!is_data_usage_cache_absent(&Error::DiskNotFound));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn data_usage_removal_maps_missing_system_volume_to_read_quorum() {
|
||||
for error in [Error::VolumeNotFound, Error::BucketNotFound(RUSTFS_META_BUCKET.to_string())] {
|
||||
assert_eq!(
|
||||
map_data_usage_metadata_read_error(error, "bucket-metadata/.usage.json"),
|
||||
Error::InsufficientReadQuorum(RUSTFS_META_BUCKET.to_string(), "bucket-metadata/.usage.json".to_string())
|
||||
);
|
||||
}
|
||||
|
||||
let missing_object = Error::ObjectNotFound(RUSTFS_META_BUCKET.to_string(), "bucket-metadata/.usage.json".to_string());
|
||||
assert_eq!(
|
||||
map_data_usage_metadata_read_error(missing_object.clone(), "bucket-metadata/.usage.json"),
|
||||
missing_object
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn load_data_usage_cache_treats_absence_as_an_empty_cache_without_retrying() {
|
||||
let name = "usage-cache";
|
||||
@@ -2872,6 +2911,22 @@ mod tests {
|
||||
);
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn data_usage_removal_surfaces_missing_system_volume_as_read_quorum() {
|
||||
for cause in [Error::BucketNotFound(RUSTFS_META_BUCKET.to_string()), Error::VolumeNotFound] {
|
||||
let store = UsageCacheReadStore::with_terminal_error(cause);
|
||||
|
||||
let error = load_data_usage_for_bucket_removal(&store, "bucket-metadata/.usage.json")
|
||||
.await
|
||||
.expect_err("missing system metadata volume must not be treated as an absent usage object");
|
||||
|
||||
assert_eq!(
|
||||
error,
|
||||
Error::InsufficientReadQuorum(RUSTFS_META_BUCKET.to_string(), "bucket-metadata/.usage.json".to_string())
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
#[tokio::test]
|
||||
async fn load_data_usage_cache_retries_a_transient_failure() {
|
||||
let name = "usage-cache";
|
||||
|
||||
Reference in New Issue
Block a user