gian/cubefs
1
0
Fork 0
mirror of https://github.com/cubefs/cubefs.git synced 2026-08-02 02:00:56 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

feat(data): if IO errors make disk status change to unavailable, not set all datapartitions of the disk as unavailable except the datapartition related to the if IO errors.

Browse Source 
Signed-off-by: true1064 <tangjingyu@oppo.com>
This commit is contained in:
true1064 2024-01-09 19:46:43 +08:00 committed by Whale Tang
parent 604e1298ee
commit cae5f00182
3 changed files with 8 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
datanode/disk.go
View File

@ -421,7 +421,7 @@ func (d *Disk) CheckDiskError(err error, rwFlag uint8) {
func (d *Disk) doDiskError() {
d.Status = proto.Unavailable
d.ForceExitRaftStore()
//d.ForceExitRaftStore()
}
func (d *Disk) triggerDiskError(rwFlag uint8, dpId uint64) {
@ -442,10 +442,10 @@ func (d *Disk) triggerDiskError(rwFlag uint8, dpId uint64) {
diskErrCnt := d.getTotalErrCnt()
diskErrPartitionCnt := d.GetDiskErrPartitionCount()
if diskErrCnt >= d.dataNode.diskUnavailableErrorCount || diskErrPartitionCnt >= d.dataNode.diskUnavailablePartitionErrorCount {
if diskErrPartitionCnt >= d.dataNode.diskUnavailablePartitionErrorCount {
msg := fmt.Sprintf("set disk unavailable for too many disk error, "+
"disk path(%v), ip(%v), diskErrCnt(%v) threshold(%v), diskErrPartitionCnt(%v) threshold(%v)",
d.Path, LocalIP, diskErrCnt, d.dataNode.diskUnavailableErrorCount, diskErrPartitionCnt, d.dataNode.diskUnavailablePartitionErrorCount)
"disk path(%v), ip(%v), diskErrCnt(%v), diskErrPartitionCnt(%v) threshold(%v)",
d.Path, LocalIP, diskErrCnt, diskErrPartitionCnt, d.dataNode.diskUnavailablePartitionErrorCount)
exporter.Warning(msg)
log.LogWarnf(msg)
d.doDiskError()
@ -462,7 +462,7 @@ func (d *Disk) updateSpaceInfo() (err error) {
mesg := fmt.Sprintf("disk path %v error on %v", d.Path, LocalIP)
log.LogErrorf(mesg)
exporter.Warning(mesg)
d.ForceExitRaftStore()
//d.ForceExitRaftStore()
} else if d.Available <= 0 {
d.Status = proto.ReadOnly
} else {

6
datanode/partition.go
View File

@ -600,9 +600,9 @@ func (dp *DataPartition) statusUpdate() {
dp.partitionStatus = proto.Unavailable
}
log.LogInfof("action[statusUpdate] dp %v raft status %v dp.status %v, status %v, disk status %v, res:%v",
dp.partitionID, dp.raftStatus, dp.Status(), status, float64(dp.disk.Status), int(math.Min(float64(status), float64(dp.disk.Status))))
dp.partitionStatus = int(math.Min(float64(status), float64(dp.disk.Status)))
log.LogInfof("action[statusUpdate] dp %v raft status %v dp.status %v, status %v, disk status %v",
dp.partitionID, dp.raftStatus, dp.Status(), status, float64(dp.disk.Status))
//dp.partitionStatus = int(math.Min(float64(status), float64(dp.disk.Status)))
}
func (dp *DataPartition) computeUsage() {

12
datanode/server.go
View File

@ -119,8 +119,6 @@ const (
ConfigServiceIDKey = "serviceIDKey"
// disk status becomes unavailable if disk error count reaches this value
ConfigKeyDiskUnavailableErrorCount = "diskUnavailableErrorCount"
// disk status becomes unavailable if disk error partition count reaches this value
ConfigKeyDiskUnavailablePartitionErrorCount = "diskUnavailablePartitionErrorCount"
)
@ -181,7 +179,6 @@ type DataNode struct {
cpuUtil atomicutil.Float64
cpuSamplerDone chan struct{}
diskUnavailableErrorCount uint64 // disk status becomes unavailable when disk error count reaches this value
diskUnavailablePartitionErrorCount uint64 // disk status becomes unavailable when disk error partition count reaches this value
}
@ -351,15 +348,6 @@ func (s *DataNode) parseConfig(cfg *config.Config) (err error) {
s.serviceIDKey = cfg.GetString(ConfigServiceIDKey)
diskUnavailableErrorCount := cfg.GetInt64(ConfigKeyDiskUnavailableErrorCount)
if diskUnavailableErrorCount <= 0 || diskUnavailableErrorCount > 100 {
diskUnavailableErrorCount = DefaultDiskUnavailableErrorCount
log.LogDebugf("action[parseConfig] ConfigKeyDiskUnavailableErrorCount(%v) out of range, set as default(%v)",
diskUnavailableErrorCount, DefaultDiskUnavailableErrorCount)
}
s.diskUnavailableErrorCount = uint64(diskUnavailableErrorCount)
log.LogDebugf("action[parseConfig] load diskUnavailableErrorCount(%v)", s.diskUnavailableErrorCount)
diskUnavailablePartitionErrorCount := cfg.GetInt64(ConfigKeyDiskUnavailablePartitionErrorCount)
if diskUnavailablePartitionErrorCount <= 0 || diskUnavailablePartitionErrorCount > 100 {
diskUnavailablePartitionErrorCount = DefaultDiskUnavailablePartitionErrorCount