mirror of
https://github.com/cubefs/cubefs.git
synced 2026-08-02 02:00:56 +00:00
453 lines
15 KiB
Go
453 lines
15 KiB
Go
// Copyright 2023 The CubeFS Authors.
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
// you may not use this file except in compliance with the License.
|
|
// You may obtain a copy of the License at
|
|
//
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
//
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
|
|
// implied. See the License for the specific language governing
|
|
// permissions and limitations under the License.
|
|
|
|
package flashnode
|
|
|
|
import (
|
|
"bytes"
|
|
"context"
|
|
"encoding/json"
|
|
"fmt"
|
|
"net"
|
|
"sync"
|
|
"time"
|
|
|
|
"github.com/cubefs/cubefs/flashnode/cachengine"
|
|
"github.com/cubefs/cubefs/proto"
|
|
"github.com/cubefs/cubefs/util"
|
|
"github.com/cubefs/cubefs/util/errors"
|
|
"github.com/cubefs/cubefs/util/exporter"
|
|
"github.com/cubefs/cubefs/util/log"
|
|
"github.com/cubefs/cubefs/util/stat"
|
|
)
|
|
|
|
func (f *FlashNode) preHandle(conn net.Conn, p *proto.Packet) error {
|
|
if (p.Opcode == proto.OpFlashNodeCacheRead || p.Opcode == proto.OpFlashNodeCachePrepare) && !f.readLimiter.Allow() {
|
|
metric := exporter.NewTPCnt("NodeReqLimit")
|
|
metric.Set(nil)
|
|
err := errors.NewErrorf("%s", "flashnode read request was been limited")
|
|
log.LogWarnf("action[preHandle] %s, remote address:%s", err.Error(), conn.RemoteAddr())
|
|
return err
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (f *FlashNode) handlePacket(conn net.Conn, p *proto.Packet) (err error) {
|
|
switch p.Opcode {
|
|
case proto.OpFlashNodeHeartbeat:
|
|
err = f.opFlashNodeHeartbeat(conn, p)
|
|
case proto.OpFlashNodeCachePrepare:
|
|
err = f.opCachePrepare(conn, p)
|
|
case proto.OpFlashNodeCacheRead:
|
|
err = f.opCacheRead(conn, p)
|
|
case proto.OpFlashNodeSetReadIOLimits:
|
|
err = f.opSetReadIOLimits(conn, p)
|
|
case proto.OpFlashNodeSetWriteIOLimits:
|
|
err = f.opSetWriteIOLimits(conn, p)
|
|
default:
|
|
err = fmt.Errorf("unknown Opcode:%d", p.Opcode)
|
|
}
|
|
|
|
return
|
|
}
|
|
|
|
func (f *FlashNode) SetTimeout(handleReadTimeout int, readDataNodeTimeout int) {
|
|
if f.handleReadTimeout != handleReadTimeout && handleReadTimeout > 0 {
|
|
log.LogInfof("FlashNode set handleReadTimeout from %d to %d", f.handleReadTimeout, handleReadTimeout)
|
|
f.handleReadTimeout = handleReadTimeout
|
|
f.limitWrite.ResetIOEx(f.diskWriteIocc*len(f.disks), f.diskWriteIoFactorFlow, f.handleReadTimeout)
|
|
f.limitWrite.ResetFlow(f.diskWriteFlow)
|
|
}
|
|
f.cacheEngine.SetReadDataNodeTimeout(readDataNodeTimeout)
|
|
}
|
|
|
|
func (f *FlashNode) opFlashNodeHeartbeat(conn net.Conn, p *proto.Packet) (err error) {
|
|
data := p.Data
|
|
req := &proto.HeartBeatRequest{}
|
|
adminTask := &proto.AdminTask{
|
|
Request: req,
|
|
}
|
|
|
|
decode := json.NewDecoder(bytes.NewBuffer(data))
|
|
decode.UseNumber()
|
|
if err = decode.Decode(adminTask); err == nil {
|
|
f.SetTimeout(req.FlashNodeHandleReadTimeout, req.FlashNodeReadDataNodeTimeout)
|
|
} else {
|
|
log.LogErrorf("decode HeartBeatRequest error: %s", err.Error())
|
|
}
|
|
|
|
resp := &proto.FlashNodeHeartbeatResponse{}
|
|
resp.Stat = make([]*proto.FlashNodeDiskCacheStat, 0)
|
|
for _, cacheStat := range f.cacheEngine.GetHeartBeatCacheStat() {
|
|
stat := &proto.FlashNodeDiskCacheStat{
|
|
DataPath: cacheStat.DataPath,
|
|
Medium: cacheStat.Medium,
|
|
Total: cacheStat.Total,
|
|
MaxAlloc: cacheStat.MaxAlloc,
|
|
HasAlloc: cacheStat.HasAlloc,
|
|
FreeSpace: cacheStat.MaxAlloc - cacheStat.HasAlloc,
|
|
HitRate: cacheStat.HitRate,
|
|
Evicts: cacheStat.Evicts,
|
|
ReadRps: f.readRps,
|
|
KeyNum: cacheStat.Num,
|
|
Status: cacheStat.Status,
|
|
}
|
|
resp.Stat = append(resp.Stat, stat)
|
|
}
|
|
writeStatus := proto.FlashNodeLimiterStatus{Status: f.limitWrite.Status(true), DiskNum: len(f.disks), ReadTimeoutSec: f.handleReadTimeout}
|
|
readStatus := proto.FlashNodeLimiterStatus{Status: f.limitRead.Status(true), DiskNum: len(f.disks), ReadTimeoutSec: f.handleReadTimeout}
|
|
resp.LimiterStatus = &proto.FlashNodeLimiterStatusInfo{WriteStatus: writeStatus, ReadStatus: readStatus}
|
|
reply, err := json.Marshal(resp)
|
|
if err != nil {
|
|
p.PacketErrorWithBody(proto.OpErr, []byte(err.Error()))
|
|
return
|
|
}
|
|
p.PacketOkWithBody(reply)
|
|
|
|
if err = p.WriteToConn(conn); err != nil {
|
|
log.LogErrorf("ack master response: %s", err.Error())
|
|
return err
|
|
}
|
|
log.LogInfof("[opMasterHeartbeat] master:%s handleReadTimeout(%v) readDataNodeTimeout(%v)",
|
|
conn.RemoteAddr().String(), req.FlashNodeHandleReadTimeout, req.FlashNodeReadDataNodeTimeout)
|
|
return
|
|
}
|
|
|
|
func (f *FlashNode) opCacheRead(conn net.Conn, p *proto.Packet) (err error) {
|
|
var volume string
|
|
bgTime := stat.BeginStat()
|
|
metricOpCacheRead := exporter.NewTPCnt("opCacheRead")
|
|
defer func() {
|
|
metricOpCacheRead.SetWithLabels(err, map[string]string{exporter.Vol: volume})
|
|
stat.EndStat("FlashNode:opCacheRead", err, bgTime, 1)
|
|
}()
|
|
|
|
defer func() {
|
|
if err != nil {
|
|
log.LogWarnf("action[opCacheRead] volume:[%s], logMsg:%s", volume,
|
|
p.LogMessage(p.GetOpMsg(), conn.RemoteAddr().String(), p.StartT, err))
|
|
p.PacketErrorWithBody(proto.OpErr, ([]byte)(err.Error()))
|
|
if e := p.WriteToConn(conn); e != nil {
|
|
log.LogErrorf("action[opCacheRead] write to conn %v", e)
|
|
}
|
|
}
|
|
}()
|
|
|
|
ctx, ctxCancel := context.WithTimeout(context.Background(), time.Duration(f.handleReadTimeout)*time.Second)
|
|
defer ctxCancel()
|
|
|
|
req := new(proto.CacheReadRequest)
|
|
if err = p.UnmarshalDataPb(req); err != nil {
|
|
return
|
|
}
|
|
if req.CacheRequest == nil {
|
|
err = fmt.Errorf("no cache read request")
|
|
return
|
|
}
|
|
volume = req.CacheRequest.Volume
|
|
|
|
cr := req.CacheRequest
|
|
block, err := f.cacheEngine.GetCacheBlockForRead(volume, cr.Inode, cr.FixedFileOffset, cr.Version, req.Size_)
|
|
if err != nil {
|
|
log.LogInfof("opCacheRead: GetCacheBlockForRead failed, req(%v) err(%v)", req, err)
|
|
hitRateMap := f.cacheEngine.GetHitRate()
|
|
for dataPath, hitRate := range hitRateMap {
|
|
if hitRate < f.lowerHitRate {
|
|
log.LogWarnf("opCacheRead: flashnode %v dataPath(%v) is lower hitrate %v", f.localAddr, dataPath, hitRate)
|
|
errMetric := exporter.NewCounter("lowerHitRate")
|
|
errMetric.AddWithLabels(1, map[string]string{exporter.FlashNode: f.localAddr, exporter.Disk: dataPath, exporter.Err: "LowerHitRate"})
|
|
}
|
|
}
|
|
bgTime2 := stat.BeginStat()
|
|
missCacheMetric := exporter.NewTPCnt("MissCacheRead")
|
|
// try to cache more miss data, but reply to client more quickly
|
|
if writable := f.limitWrite.TryRunWithContext(ctx, int(req.Size_), func() {
|
|
if block2, err := f.cacheEngine.CreateBlock(cr, conn.RemoteAddr().String(), false); err != nil {
|
|
log.LogWarnf("opCacheRead: CreateBlock failed, req(%v) err(%v)", req, err)
|
|
return
|
|
} else {
|
|
block2.InitOnceForCacheRead(f.cacheEngine, cr.Sources)
|
|
}
|
|
}); !writable {
|
|
err = fmt.Errorf("create block cache limited")
|
|
stat.EndStat("MissCacheReadLimit", err, bgTime2, 1)
|
|
missCacheMetric.SetWithLabels(err, map[string]string{exporter.Vol: volume})
|
|
return
|
|
}
|
|
select {
|
|
case <-ctx.Done():
|
|
stat.EndStat("MissCacheReadCancel", ctx.Err(), bgTime2, 1)
|
|
missCacheMetric.SetWithLabels(err, map[string]string{exporter.Vol: volume})
|
|
return ctx.Err()
|
|
default:
|
|
block, err = f.cacheEngine.GetCacheBlockForRead(volume, cr.Inode, cr.FixedFileOffset, cr.Version, req.Size_)
|
|
}
|
|
stat.EndStat("MissCacheRead", err, bgTime2, 1)
|
|
missCacheMetric.SetWithLabels(err, map[string]string{exporter.Vol: volume})
|
|
if err != nil {
|
|
return err
|
|
}
|
|
}
|
|
bgTime2 := stat.BeginStat()
|
|
hitCacheMetric := exporter.NewTPCnt("HitCacheRead")
|
|
// reply to client as quick as possible if hit cache
|
|
err2 := f.limitRead.RunNoWait(int(req.Size_), false, func() {
|
|
err = f.doStreamReadRequest(ctx, conn, req, p, block)
|
|
})
|
|
if err2 != nil {
|
|
err = err2
|
|
stat.EndStat("HitCacheRead", err, bgTime2, 1)
|
|
hitCacheMetric.SetWithLabels(err, map[string]string{exporter.Vol: volume})
|
|
return
|
|
}
|
|
stat.EndStat("HitCacheRead", err, bgTime2, 1)
|
|
hitCacheMetric.SetWithLabels(err, map[string]string{exporter.Vol: volume})
|
|
return
|
|
}
|
|
|
|
func (f *FlashNode) doStreamReadRequest(ctx context.Context, conn net.Conn, req *proto.CacheReadRequest, p *proto.Packet, block *cachengine.CacheBlock) (err error) {
|
|
const action = "action[doStreamReadRequest]"
|
|
needReplySize := uint32(req.Size_)
|
|
offset := int64(req.Offset)
|
|
defer func() {
|
|
if err != nil {
|
|
log.LogWarnf("%s cache block(%v) err:%v", action, block.String(), err)
|
|
} else {
|
|
f.metrics.updateReadCountMetric(block.GetRootPath())
|
|
f.metrics.updateReadBytesMetric(req.Size_, block.GetRootPath())
|
|
}
|
|
}()
|
|
for needReplySize > 0 {
|
|
err = nil
|
|
reply := proto.NewPacket()
|
|
reply.ReqID = p.ReqID
|
|
reply.StartT = p.StartT
|
|
|
|
currReadSize := uint32(util.Min(int(needReplySize), util.ReadBlockSize))
|
|
|
|
var bufOnce sync.Once
|
|
buf, bufErr := proto.Buffers.Get(util.ReadBlockSize)
|
|
bufRelease := func() {
|
|
bufOnce.Do(func() {
|
|
if bufErr == nil {
|
|
proto.Buffers.Put(reply.Data[:util.ReadBlockSize])
|
|
}
|
|
})
|
|
}
|
|
if bufErr != nil {
|
|
buf = make([]byte, currReadSize)
|
|
}
|
|
reply.Data = buf[:currReadSize]
|
|
|
|
reply.ExtentOffset = offset
|
|
p.Size = currReadSize
|
|
p.ExtentOffset = offset
|
|
reply.CRC, err = block.Read(ctx, reply.Data[:], offset, int64(currReadSize))
|
|
if err != nil {
|
|
bufRelease()
|
|
return
|
|
}
|
|
p.CRC = reply.CRC
|
|
|
|
reply.Size = currReadSize
|
|
reply.ResultCode = proto.OpOk
|
|
reply.Opcode = p.Opcode
|
|
p.ResultCode = proto.OpOk
|
|
|
|
bgTime := stat.BeginStat()
|
|
metric := exporter.NewTPCnt("HitCacheRead_ReplyToClient")
|
|
if err = reply.WriteToConn(conn); err != nil {
|
|
metric.SetWithLabels(err, map[string]string{exporter.Vol: req.CacheRequest.Volume, exporter.Client: conn.RemoteAddr().String()})
|
|
bufRelease()
|
|
log.LogErrorf("%s volume:[%s] %s", action, req.CacheRequest.Volume,
|
|
reply.LogMessage(reply.GetOpMsg(), conn.RemoteAddr().String(), reply.StartT, err))
|
|
return
|
|
}
|
|
stat.EndStat("HitCacheRead:ReplyToClient", err, bgTime, 1)
|
|
metric.SetWithLabels(err, map[string]string{exporter.Vol: req.CacheRequest.Volume, exporter.Client: conn.RemoteAddr().String()})
|
|
needReplySize -= currReadSize
|
|
offset += int64(currReadSize)
|
|
bufRelease()
|
|
if log.EnableInfo() {
|
|
log.LogInfof("%s ReqID[%d] volume:[%s] reply[%s] block[%s]", action, p.ReqID, req.CacheRequest.Volume,
|
|
reply.LogMessage(reply.GetOpMsg(), conn.RemoteAddr().String(), reply.StartT, err), block.String())
|
|
}
|
|
}
|
|
p.PacketOkReply()
|
|
return
|
|
}
|
|
|
|
func (f *FlashNode) opCachePrepare(conn net.Conn, p *proto.Packet) (err error) {
|
|
action := "action[opCachePrepare]"
|
|
var volume string
|
|
defer func() {
|
|
if err != nil {
|
|
log.LogErrorf("%s volume:[%s] %s", action, volume,
|
|
p.LogMessage(p.GetOpMsg(), conn.RemoteAddr().String(), p.StartT, err))
|
|
p.PacketErrorWithBody(proto.OpErr, ([]byte)(err.Error()))
|
|
if e := p.WriteToConn(conn); e != nil {
|
|
log.LogErrorf("%s write to conn %v", action, e)
|
|
}
|
|
}
|
|
}()
|
|
|
|
req := new(proto.CachePrepareRequest)
|
|
if err = p.UnmarshalDataPb(req); err != nil {
|
|
return
|
|
}
|
|
if req.CacheRequest == nil {
|
|
err = fmt.Errorf("no cache prepare request")
|
|
return
|
|
}
|
|
volume = req.CacheRequest.Volume
|
|
|
|
if err = f.cacheEngine.PrepareCache(p.ReqID, req.CacheRequest, conn.RemoteAddr().String()); err != nil {
|
|
log.LogErrorf("%s prepare %v", action, err)
|
|
return
|
|
}
|
|
|
|
p.PacketOkReply()
|
|
if e := p.WriteToConn(conn); e != nil {
|
|
log.LogErrorf("%s write to conn volume:%s %v", action, volume, e)
|
|
}
|
|
|
|
if len(req.FlashNodes) > 0 {
|
|
f.dispatchRequestToFollowers(req)
|
|
}
|
|
return
|
|
}
|
|
|
|
func (f *FlashNode) dispatchRequestToFollowers(request *proto.CachePrepareRequest) {
|
|
req := &proto.CachePrepareRequest{
|
|
CacheRequest: request.CacheRequest,
|
|
FlashNodes: make([]string, 0),
|
|
}
|
|
wg := sync.WaitGroup{}
|
|
for idx, addr := range request.FlashNodes {
|
|
if addr == f.localAddr {
|
|
continue
|
|
}
|
|
wg.Add(1)
|
|
log.LogDebugf("dispatchRequestToFollowers: try to prepare on addr:%s (%d/%d)", addr, idx, len(request.FlashNodes))
|
|
go func(addr string) {
|
|
defer wg.Done()
|
|
if err := f.sendPrepareRequest(addr, req); err != nil {
|
|
log.LogErrorf("dispatchRequestToFollowers: failed to distribute request to addr(%v) err(%v)", addr, err)
|
|
}
|
|
}(addr)
|
|
}
|
|
wg.Wait()
|
|
}
|
|
|
|
func (f *FlashNode) sendPrepareRequest(addr string, req *proto.CachePrepareRequest) (err error) {
|
|
action := "action[sendPrepareRequest]"
|
|
conn, err := f.connPool.GetConnect(addr)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer func() {
|
|
f.connPool.PutConnect(conn, err != nil)
|
|
}()
|
|
log.LogDebugf("%s to addr:%s request:%v", action, addr, req)
|
|
|
|
followerPacket := proto.NewPacketReqID()
|
|
followerPacket.Opcode = proto.OpFlashNodeCachePrepare
|
|
if err = followerPacket.MarshalDataPb(req); err != nil {
|
|
log.LogWarnf("%s failed to MarshalDataPb (%+v) err(%v)", action, followerPacket, err)
|
|
return err
|
|
}
|
|
if err = followerPacket.WriteToNoDeadLineConn(conn); err != nil {
|
|
log.LogWarnf("%s failed to write to addr(%s) err(%v)", action, addr, err)
|
|
return err
|
|
}
|
|
reply := proto.NewPacket()
|
|
if err = reply.ReadFromConn(conn, 30); err != nil {
|
|
log.LogWarnf("%s read reply(%v) from addr(%s) err(%v)", action, reply, addr, err)
|
|
return err
|
|
}
|
|
if reply.ResultCode != proto.OpOk {
|
|
log.LogWarnf("%s reply(%v) from addr(%s) ResultCode(%d)", action, reply, addr, reply.ResultCode)
|
|
return fmt.Errorf("ResultCode(%v)", reply.ResultCode)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (f *FlashNode) opSetReadIOLimits(conn net.Conn, p *proto.Packet) (err error) {
|
|
data := p.Data
|
|
req := &proto.FlashNodeSetIOLimitsRequest{}
|
|
adminTask := &proto.AdminTask{
|
|
Request: req,
|
|
}
|
|
update := false
|
|
decode := json.NewDecoder(bytes.NewBuffer(data))
|
|
decode.UseNumber()
|
|
if err = decode.Decode(adminTask); err == nil {
|
|
log.LogDebugf("opSetReadIOLimits req: %v", req)
|
|
if req.Iocc != -1 {
|
|
f.diskReadIocc = req.Iocc
|
|
update = true
|
|
}
|
|
if req.Flow != -1 {
|
|
f.diskReadFlow = req.Flow
|
|
update = true
|
|
}
|
|
if req.Factor != -1 {
|
|
f.diskReadIoFactorFlow = req.Factor
|
|
update = true
|
|
}
|
|
if update {
|
|
f.limitRead.ResetIOEx(f.diskReadIocc*len(f.disks), f.diskReadIoFactorFlow, f.handleReadTimeout)
|
|
f.limitRead.ResetFlow(f.diskReadFlow)
|
|
}
|
|
} else {
|
|
log.LogErrorf("decode FlashNodeSetIOLimitsRequest error: %s", err.Error())
|
|
}
|
|
p.PacketOkReply()
|
|
return
|
|
}
|
|
|
|
func (f *FlashNode) opSetWriteIOLimits(conn net.Conn, p *proto.Packet) (err error) {
|
|
data := p.Data
|
|
req := &proto.FlashNodeSetIOLimitsRequest{}
|
|
adminTask := &proto.AdminTask{
|
|
Request: req,
|
|
}
|
|
update := false
|
|
decode := json.NewDecoder(bytes.NewBuffer(data))
|
|
decode.UseNumber()
|
|
if err = decode.Decode(adminTask); err == nil {
|
|
log.LogDebugf("opSetReadIOLimits req: %v", req)
|
|
if req.Iocc != -1 {
|
|
f.diskWriteIocc = req.Iocc
|
|
update = true
|
|
}
|
|
if req.Flow != -1 {
|
|
f.diskWriteFlow = req.Flow
|
|
update = true
|
|
}
|
|
if req.Factor != -1 {
|
|
f.diskWriteIoFactorFlow = req.Factor
|
|
update = true
|
|
}
|
|
if update {
|
|
f.limitWrite.ResetIOEx(f.diskWriteIocc*len(f.disks), f.diskWriteIoFactorFlow, f.handleReadTimeout)
|
|
f.limitWrite.ResetFlow(f.diskWriteFlow)
|
|
}
|
|
} else {
|
|
log.LogErrorf("decode opSetWriteIOLimits error: %s", err.Error())
|
|
}
|
|
p.PacketOkReply()
|
|
return
|
|
}
|
