Files
seaweedfs/weed/server/volume_server.go
T
e3e02d3364 [CheckDisk]: implement disk health detection (#9560)
* [CheckDisk][GRPC]: implement MVP for disk health detection, added timeout for new grpc connections

* fix(volume): build disk health check on every platform

setDiskStatus only existed behind the statfs build tag, so disk.go failed
to compile on windows, openbsd, solaris, netbsd and plan9. Move the timeout
wrapper and failure tracking into the shared disk.go and have each platform's
fillInDiskStatus return an error, so every platform gets the same protection
from a stuck filesystem.

Also restore the uint64(fs.Bavail) cast: Bavail is int64 on freebsd, so the
unguarded multiply broke the freebsd build.

* fix(volume): keep one outstanding statfs probe per disk

A stuck statfs used to leave isChecking cleared by the timeout path, so the
next check spawned another goroutine while the previous one was still blocked
in the syscall, leaking one goroutine per minute on a hung disk. Clear the
flag only when statfs returns and treat an overlapping check as a failure, so
a hung filesystem keeps a single outstanding probe and still gets reported.

* fix(volume): assume disk available until the first health check

isDiskAvailable defaulted to false, and CollectHeartbeat skips locations that
are not available. A freshly started volume server would therefore omit every
volume from its first heartbeats until the async CheckDiskSpace ran, so the
master could briefly treat all of them as missing.

* fix(volume): label the disk error metric by data directory

The new gauge tagged the series with IdxDirectory while every neighbouring
resource gauge uses Directory, so the error series would not line up with them
in dashboards. Also log the underlying error instead of a generic message.

* test(volume): cover disk health success and repeated-failure paths

* fix(volume): make a healthy disk the zero-value default

Track the disk as isDiskUnavailable instead of isDiskAvailable so the safe
state is the zero value, matching isDiskSpaceLow. CollectHeartbeat only skips a
location once a check has actively marked it unavailable, so any DiskLocation
built without running CheckDiskSpace (tests, future call sites) still reports
its volumes instead of silently dropping them.

* feat(disk): detect degraded disks using IO latency probes

* feat(stats): introduce configurable disk I/O health probe with EWMA-based latency detection

* feat(disk): replace EWMA with sliding window algorithm for disk health detection and added user-friendly options

* feat(disk): improve disk health probing and recovery

* feat(volume): configure disk health checks via volume.toml

* fix(volume): Remove disk IO probe CLI options

---------

Co-authored-by: ptukha <ptukha@tochka.com>
Co-authored-by: Chris Lu <chris.lu@gmail.com>
2026-06-02 09:02:05 -07:00

210 lines
7.6 KiB
Go

package weed_server
import (
"fmt"
"net/http"
"sync"
"time"
"github.com/seaweedfs/seaweedfs/weed/pb"
"github.com/seaweedfs/seaweedfs/weed/pb/volume_server_pb"
"github.com/seaweedfs/seaweedfs/weed/storage/types"
"google.golang.org/grpc"
"github.com/seaweedfs/seaweedfs/weed/stats"
"github.com/seaweedfs/seaweedfs/weed/util"
"github.com/seaweedfs/seaweedfs/weed/glog"
"github.com/seaweedfs/seaweedfs/weed/security"
"github.com/seaweedfs/seaweedfs/weed/storage"
)
type VolumeServer struct {
volume_server_pb.UnimplementedVolumeServerServer
inFlightUploadDataSize int64
inFlightDownloadDataSize int64
concurrentUploadLimit int64
concurrentDownloadLimit int64
inFlightUploadDataLimitCond *sync.Cond
inFlightDownloadDataLimitCond *sync.Cond
inflightUploadDataTimeout time.Duration
inflightDownloadDataTimeout time.Duration
hasSlowRead bool
readBufferSizeMB int
SeedMasterNodes []pb.ServerAddress
// seedMasterSet mirrors SeedMasterNodes keyed by the canonical http
// form. It is computed once in NewVolumeServer so admission paths can
// answer is-this-a-seed-master in O(1).
seedMasterSet map[string]struct{}
whiteList []string
currentMaster pb.ServerAddress
currentMasterLock sync.RWMutex
pulsePeriod time.Duration
dataCenter string
rack string
store *storage.Store
guard *security.Guard
grpcDialOption grpc.DialOption
needleMapKind storage.NeedleMapKind
ldbTimout int64
FixJpgOrientation bool
ReadMode string
AllowUntrustedRemoteEndpoints bool
compactionBytePerSecond int64
maintenanceBytePerSecond int64
metricsAddress string
metricsIntervalSec int
fileSizeLimitBytes int64
isHeartbeating bool
stopChan chan bool
}
func NewVolumeServer(adminMux, publicMux *http.ServeMux, ip string,
port int, grpcPort int, publicUrl string, id string,
folders []string, maxCounts []int32, minFreeSpaces []util.MinFreeSpace, diskTypes []types.DiskType, diskTags [][]string,
idxFolder string,
needleMapKind storage.NeedleMapKind,
masterNodes []pb.ServerAddress, pulsePeriod time.Duration,
dataCenter string, rack string,
whiteList []string,
fixJpgOrientation bool,
readMode string,
compactionMBPerSecond int,
maintenanceMBPerSecond int,
fileSizeLimitMB int,
concurrentUploadLimit int64,
concurrentDownloadLimit int64,
inflightUploadDataTimeout time.Duration,
inflightDownloadDataTimeout time.Duration,
hasSlowRead bool,
readBufferSizeMB int,
ldbTimeout int64,
allowUntrustedRemoteEndpoints bool,
diskProbeConfig stats.DiskIOProbeConfig,
) *VolumeServer {
v := util.GetViper()
signingKey := v.GetString("jwt.signing.key")
v.SetDefault("jwt.signing.expires_after_seconds", 10)
expiresAfterSec := v.GetInt("jwt.signing.expires_after_seconds")
enableUiAccess := v.GetBool("access.ui")
readSigningKey := v.GetString("jwt.signing.read.key")
v.SetDefault("jwt.signing.read.expires_after_seconds", 60)
readExpiresAfterSec := v.GetInt("jwt.signing.read.expires_after_seconds")
vs := &VolumeServer{
pulsePeriod: pulsePeriod,
dataCenter: dataCenter,
rack: rack,
needleMapKind: needleMapKind,
FixJpgOrientation: fixJpgOrientation,
ReadMode: readMode,
grpcDialOption: security.LoadClientTLS(util.GetViper(), "grpc.volume"),
compactionBytePerSecond: int64(compactionMBPerSecond) * 1024 * 1024,
maintenanceBytePerSecond: int64(maintenanceMBPerSecond) * 1024 * 1024,
fileSizeLimitBytes: int64(fileSizeLimitMB) * 1024 * 1024,
isHeartbeating: true,
stopChan: make(chan bool),
inFlightUploadDataLimitCond: sync.NewCond(new(sync.Mutex)),
inFlightDownloadDataLimitCond: sync.NewCond(new(sync.Mutex)),
concurrentUploadLimit: concurrentUploadLimit,
concurrentDownloadLimit: concurrentDownloadLimit,
inflightUploadDataTimeout: inflightUploadDataTimeout,
inflightDownloadDataTimeout: inflightDownloadDataTimeout,
hasSlowRead: hasSlowRead,
readBufferSizeMB: readBufferSizeMB,
ldbTimout: ldbTimeout,
whiteList: whiteList,
AllowUntrustedRemoteEndpoints: allowUntrustedRemoteEndpoints,
}
whiteList = append(whiteList, util.StringSplit(v.GetString("guard.white_list"), ",")...)
// Copy the caller's slice so subsequent external mutation cannot desync
// SeedMasterNodes from the frozen lookup set built below.
seedMasters := make([]pb.ServerAddress, len(masterNodes))
copy(seedMasters, masterNodes)
vs.SeedMasterNodes = seedMasters
vs.seedMasterSet = make(map[string]struct{}, len(seedMasters))
for _, m := range seedMasters {
vs.seedMasterSet[m.ToHttpAddress()] = struct{}{}
}
vs.checkWithMaster()
vs.store = storage.NewStore(vs.grpcDialOption, ip, port, grpcPort, publicUrl, id, folders, maxCounts, minFreeSpaces, idxFolder, vs.needleMapKind, diskTypes, diskTags, ldbTimeout, diskProbeConfig)
vs.guard = security.NewGuard(whiteList, signingKey, expiresAfterSec, readSigningKey, readExpiresAfterSec)
handleStaticResources(adminMux)
adminMux.HandleFunc("/status", requestIDMiddleware(vs.statusHandler))
adminMux.HandleFunc("/healthz", requestIDMiddleware(vs.healthzHandler))
adminMux.HandleFunc("/readyz", requestIDMiddleware(vs.healthzHandler))
if signingKey == "" || enableUiAccess {
// only expose the volume server details for safe environments
adminMux.HandleFunc("/ui/index.html", requestIDMiddleware(vs.uiStatusHandler))
/*
adminMux.HandleFunc("/stats/counter", vs.guard.WhiteList(statsCounterHandler))
adminMux.HandleFunc("/stats/memory", vs.guard.WhiteList(statsMemoryHandler))
adminMux.HandleFunc("/stats/disk", vs.guard.WhiteList(vs.statsDiskHandler))
*/
}
adminMux.HandleFunc("/", requestIDMiddleware(vs.privateStoreHandler))
if publicMux != adminMux {
// separated admin and public port
handleStaticResources(publicMux)
publicMux.HandleFunc("/", requestIDMiddleware(vs.publicReadOnlyHandler))
}
stats.VolumeServerConcurrentDownloadLimit.Set(float64(vs.concurrentDownloadLimit))
stats.VolumeServerConcurrentUploadLimit.Set(float64(vs.concurrentUploadLimit))
stats.VolumeServerStartTimeSeconds.Set(float64(time.Now().Unix()))
go vs.heartbeat()
go stats.LoopPushingMetric("volumeServer", util.JoinHostPort(ip, port), vs.metricsAddress, vs.metricsIntervalSec)
return vs
}
func (vs *VolumeServer) SetStopping() {
glog.V(0).Infoln("Stopping volume server...")
vs.store.SetStopping()
}
func (vs *VolumeServer) LoadNewVolumes() {
glog.V(0).Infoln(" Loading new volume ids ...")
vs.store.LoadNewVolumes()
}
func (vs *VolumeServer) Shutdown() {
glog.V(0).Infoln("Shutting down volume server...")
vs.store.Close()
glog.V(0).Infoln("Shut down successfully!")
}
func (vs *VolumeServer) Reload() {
glog.V(0).Infoln("Reload volume server...")
util.LoadConfiguration("security", false)
v := util.GetViper()
vs.guard.UpdateWhiteList(append(vs.whiteList, util.StringSplit(v.GetString("guard.white_list"), ",")...))
}
// Returns whether a volume server is in maintenance (i.e. read-only) mode.
func (vs *VolumeServer) MaintenanceMode() bool {
if vs.store == nil {
return false
}
return vs.store.State.Proto().GetMaintenance()
}
// Checks if a volume server is in maintenance mode, and returns an error explaining why.
func (vs *VolumeServer) CheckMaintenanceMode() error {
if !vs.MaintenanceMode() {
return nil
}
return fmt.Errorf("volume server %s is in maintenance mode", vs.store.Id)
}