gsadmin/rustfs
1
0
Fork 0
mirror of https://github.com/rustfs/rustfs.git synced 2026-08-19 19:16:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(heal): aggregate status across cluster nodes (#4990)

Browse Source 
* fix(rpc): bind internode auth to exact targets

* fix(heal): initialize the runtime atomically

* fix(heal): aggregate status across cluster nodes

---------

Co-authored-by: Zhengchao An <anzhengchao@gmail.com>
This commit is contained in:
cxymds
2026-07-19 23:25:52 +08:00
committed by GitHub
parent 056ebcee38
commit 4290f390dd
10 changed files with 1494 additions and 230 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crates/ecstore/src/cluster/rpc/peer_rest_client.rs
+38 -5
View File
@@ -29,11 +29,11 @@ use rustfs_madmin::{
};
use rustfs_protos::evict_failed_connection;
use rustfs_protos::proto_gen::node_service::{
CancelDecommissionRequest, ClearDecommissionRequest, DeleteBucketMetadataRequest, DeletePolicyRequest,
DeleteServiceAccountRequest, DeleteUserRequest, GetCpusRequest, GetLiveEventsRequest, GetMemInfoRequest, GetMetricsRequest,
GetNetInfoRequest, GetOsInfoRequest, GetPartitionsRequest, GetProcInfoRequest, GetSeLinuxInfoRequest, GetSysConfigRequest,
GetSysErrorsRequest, LoadBucketMetadataRequest, LoadGroupRequest, LoadPolicyMappingRequest, LoadPolicyRequest,
LoadRebalanceMetaRequest, LoadServiceAccountRequest, LoadTransitionTierConfigRequest, LoadUserRequest,
BackgroundHealStatusRequest, CancelDecommissionRequest, ClearDecommissionRequest, DeleteBucketMetadataRequest,
DeletePolicyRequest, DeleteServiceAccountRequest, DeleteUserRequest, GetCpusRequest, GetLiveEventsRequest, GetMemInfoRequest,
GetMetricsRequest, GetNetInfoRequest, GetOsInfoRequest, GetPartitionsRequest, GetProcInfoRequest, GetSeLinuxInfoRequest,
GetSysConfigRequest, GetSysErrorsRequest, LoadBucketMetadataRequest, LoadGroupRequest, LoadPolicyMappingRequest,
LoadPolicyRequest, LoadRebalanceMetaRequest, LoadServiceAccountRequest, LoadTransitionTierConfigRequest, LoadUserRequest,
LocalStorageInfoRequest, Mss, ReloadPoolMetaRequest, ReloadSiteReplicationConfigRequest, ScannerActivityRequest,
ScannerActivityResponse, ServerInfoRequest, SignalServiceRequest, StartDecommissionRequest, StartProfilingRequest,
StopRebalanceRequest, node_service_client::NodeServiceClient,
@@ -60,6 +60,7 @@ pub const PEER_RESTSUB_SYS: &str = "sub-sys";
pub const PEER_RESTDRY_RUN: &str = "dry-run";
pub const SERVICE_SIGNAL_REFRESH_CONFIG: u64 = 1;
pub const SERVICE_SIGNAL_RELOAD_DYNAMIC: u64 = 2;
const BACKGROUND_HEAL_STATUS_MAX_MESSAGE_SIZE: usize = 64 * 1024;
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;
@@ -664,6 +665,38 @@ impl PeerRestClient {
Err(Error::NotImplemented)
}
pub async fn background_heal_status(&self) -> Result<Option<Vec<u8>>> {
self.finalize_result(
async {
let mut client = self
.get_client()
.await?
.max_decoding_message_size(BACKGROUND_HEAL_STATUS_MAX_MESSAGE_SIZE);
let response = match client
.background_heal_status(Request::new(BackgroundHealStatusRequest::default()))
.await
{
Ok(response) => response.into_inner(),
Err(status) if status.code() == tonic::Code::Unimplemented => {
// RUSTFS_COMPAT_TODO(heal-status-rpc-v1): accept old peers without node heal snapshots. Remove after the minimum supported RustFS peer version implements BackgroundHealStatus.
return Ok(None);
}
Err(status) => return Err(status.into()),
};
if !response.success {
return Err(Error::other(
response
.error_info
.unwrap_or_else(|| "peer background heal status failed without an error".to_string()),
));
}
Ok(Some(response.bg_heal_state.to_vec()))
}
.await,
)
.await
}
pub async fn load_bucket_metadata(&self, bucket: &str, scanner_maintenance_change: bool) -> Result<()> {
self.finalize_result(
async {