unstable->3.5 (#2397)

* fix: codis-dashboard uses 100% cpu(#2332) (#2393) Co-authored-by: liuchengyu <[email protected]> * fix: The role displayed on the first Server in the Group area of the codis-fe is incorrect (#2350) (#2387) Co-authored-by: liuchengyu <[email protected]> * fix: automatic fix master-slave replication relationship after master or slave service restarted (#2373, #2038, #1950, #1967, #2351)) (#2386) Co-authored-by: liuchengyu <[email protected]> * feat:add 3.5.3 changelog (#2395) * add 3.5.3 changelog --------- Co-authored-by: chejinge <[email protected]> --------- Co-authored-by: Chengyu Liu <[email protected]> Co-authored-by: liuchengyu <[email protected]> Co-authored-by: chejinge <[email protected]>
OpenAtomFoundation · Feb 7, 2024 · f61f49a · f61f49a
1 parent e93028f
commit f61f49a
Show file tree

Hide file tree

Showing 15 changed files with 658 additions and 277 deletions.
diff --git a/CHANGELOG.MD b/CHANGELOG.MD
@@ -1,3 +1,92 @@
+# v3.5.3
+
+## New features
+
+- Pika supports ACL[#2013](https://github.com/OpenAtomFoundation/pika/pull/2013) @[lqxhub](https://github.com/lqxhub)
+
+- Automatically resume service when Codis dashboard coroutine panics[#2349](https://github.com/OpenAtomFoundation/pika/pull/2349)@[chengyu-l](https://github.com/chengyu-l)
+
+- During the full replication process, the slave node of the pika service does not receive read traffic requests.[#2197](https://github.com/OpenAtomFoundation/pika/pull/2197) @[tedli](https://github.com/tedli)
+
+- Pika cache adds bimap data type.[#2197](https://github.com/OpenAtomFoundation/pika/pull/2197) @[chejinge](https://github.com/chejinge)
+
+- Delete the remaining Slots in Sharing mode. There is only DB under Pika, and there are multiple DBs under one Pika.[#2251](https://github.com/OpenAtomFoundation/pika/pull/2251) @[Mixficsol](https://github.com/Mixficsol)
+
+- Pika exporter exposes cache-related data collection indicators.[#2318](https://github.com/OpenAtomFoundation/pika/pull/2318) @[dingxiaoshuai](https://github.com/dingxiaoshuai123)
+
+- Pika supports separation of fast and slow commands.[#2162](https://github.com/OpenAtomFoundation/pika/pull/2162) @[dingxiaoshuai](https://github.com/dingxiaoshuai123)
+
+- After pika executes bgsave, retain the unix timepoint.[#2167](https://github.com/OpenAtomFoundation/pika/pull/2167) @[hero-heng](https://github.com/hero-heng)
+
+- Pika supports dynamic configuration of the disable_auto_compations parameter.[#2257](https://github.com/OpenAtomFoundation/pika/pull/2257) @[hero-heng](https://github.com/hero-heng)
+
+- Pika supports Redis Stream.[#1955](https://github.com/OpenAtomFoundation/pika/pull/1955) @[KKorpse](https://github.com/KKorpse)
+
+- Pika supports large key analysis tools[#2195](https://github.com/OpenAtomFoundation/pika/pull/2195) @[sjcsjc123](https://github.com/sjcsjc123)
+
+- Pika supports dynamic adjustment of Pika cache parameters[#2197](https://github.com/OpenAtomFoundation/pika/pull/2197) @[chejinge](https://github.com/chejinge)
+
+- Updated Pika benchmark tool to support more interface stress tests.[#2222](https://github.com/OpenAtomFoundation/pika/pull/2222)@[wangshao1](https://github.com/wangshao1)
+
+- Pika Operator supports automatic expansion of pika clusters.[#2121](https://github.com/OpenAtomFoundation/pika/pull/2121)@[machinly](https://github.com/machinly/)
+
+- Add the CompactRange command to support compacting keys within a certain range.[#2163](https://github.com/OpenAtomFoundation/pika/pull/2163)@[u6th9d](https://github.com/u6th9d)
+
+- Add small time cost compaction policy.[#2172](https://github.com/OpenAtomFoundation/pika/pull/2172)@[u6th9d](https://github.com/u6th9d)
+
+- Upgrade RocksDB version to v8.7.3.[#2157](https://github.com/OpenAtomFoundation/pika/pull/2157)@[JasirVoriya](https://github.com/JasirVoriya)
+
+- Pika distributed cluster Codis proxy adds new observable indicators.[#2199](https://github.com/OpenAtomFoundation/pika/pull/2199)@[dingxiaoshuai](https://github.com/dingxiaoshuai123)
+
+- Pika distributed cluster supports automatic failover.[#2386](https://github.com/OpenAtomFoundation/pika/pull/2386)@[chengyu-l](https://github.com/chengyu-l)
+
+## bugfix
+
+- Fixed an issue where Pika would accidentally delete dump files during full replication from the node.[#2377](https://github.com/OpenAtomFoundation/pika/pull/2377)@[wangshao1](https://github.com/wangshao1)
+
+- Fixed the processing logic after the slave node receives an abnormal response packet from the master during the master-slave replication process.[#2319](https://github.com/OpenAtomFoundation/pika/pull/2319)@[wangshao1](https://github.com/wangshao1)
+
+- Call disable compaction when pika executes the shutdown command to improve the process exit speed. [#2345](https://github.com/OpenAtomFoundation/pika/pull/2345) @[panlei-coder](https://github.com/panlei-coder)
+
+- Fix the problem of inaccurate Codis-dashboard Redis Memory value.[#2337](https://github.com/OpenAtomFoundation/pika/pull/2337) @[Mixficsol](https://github.com/Mixficsol)
+
+- The INFO command is time-consuming and optimized to reduce the frequency of disk checks. [#2197](https://github.com/OpenAtomFoundation/pika/pull/2197) @[chejinge](https://github.com/chejinge)
+
+- Fixed the issue where rsync deletes temporary files with incorrect paths and fails to delete them, causing rocksdb to fail to open.[#2186](https://github.com/OpenAtomFoundation/pika/pull/2186)@[wangshao1](https://github.com/wangshao1)
+
+- Fixed the problem that the compact, bgsave, and info keyspace commands did not specify the db name, resulting in some coredump commands.[#2194](https://github.com/OpenAtomFoundation/pika/pull/2194)@[u6th9d](https://github.com/u6th9d)
+
+- Codis dashboard uses info replication instead of info command to search master ip to reduce the performance impact on Pika. [#2198](https://github.com/OpenAtomFoundation/pika/pull/2198) @[chenbt-hz](https://github.com/chenbt-hz)
+
+- Fix Pika cache to use edge cases to solve the problem of cache and DB data inconsistency in some scenarios.[#2225](https://github.com/OpenAtomFoundation/pika/pull/2225) @[chejinge](https://github.com/chejinge)
+
+- Fixed the issue where Segmentation fault would be reported when the dump folder is empty.[#2265](https://github.com/OpenAtomFoundation/pika/pull/2265) @[chenbt-hz](https://github.com/chenbt-hz)
+
+- Fixed the problem that some command caches did not take effect due to flag calculation errors.[#2217](https://github.com/OpenAtomFoundation/pika/pull/2217) @[lqxhub](https://github.com/lqxhub)
+
+- Fixed the problem that in master-slave replication mode, after the master instance flushdb, the slave instance cannot be accessed due to deadlock.[#2249](https://github.com/OpenAtomFoundation/pika/pull/2249)@[ForestLH](https://github.com/ForestLH)
+
+- Fixed the issue where some commands did not judge the return value of RocksDB.[#2187](https://github.com/OpenAtomFoundation/pika/pull/2187)@[callme-taota](https://github.com/callme-taota)
+
+- Fixed the problem that some command caches did not take effect due to flag calculation errors.[#2217](https://github.com/OpenAtomFoundation/pika/pull/2217) @[lqxhub](https://github.com/lqxhub)
+
+- Fixed the problem that in master-slave replication mode, after the master instance flushdb, the slave instance cannot be accessed due to deadlock.[#2249](https://github.com/OpenAtomFoundation/pika/pull/2249)@[ForestLH](https://github.com/ForestLH)
+
+- Fixed the issue where some commands did not judge the return value of RocksDB.[#2187](https://github.com/OpenAtomFoundation/pika/pull/2187)@[callme-taota](https://github.com/callme-taota)
+
+- Fix the problem of info keyspace returning wrong results.[#2369](https://github.com/OpenAtomFoundation/pika/pull/2369)@[Mixficsol](https://github.com/Mixficsol)
+
+- Standard function return value and initial value.[#2176](https://github.com/OpenAtomFoundation/pika/pull/2176)@[Mixficsol](https://github.com/Mixficsol)
+
+- Fixed the problem of inaccurate network monitoring indicator statistics.[#2234](https://github.com/OpenAtomFoundation/pika/pull/2234)@[chengyu-l](https://github.com/chengyu-l)
+
+- Fixed an issue where some parameters in configuration file loading were abnormal.[#2218](https://github.com/OpenAtomFoundation/pika/pull/2218)@[jettcc](https://github.com/jettcc)
+
+- Fix Codis dashboard cpu used 100%.[#2393](https://github.com/OpenAtomFoundation/pika/pull/2393)@[chengyu-l](https://github.com/chengyu-l)
+
+- Fix the problem of abnormal display of master and slave roles in Codis fe of pika.[#2387](https://github.com/OpenAtomFoundation/pika/pull/2387)@[chengyu-l](https://github.com/chengyu-l)
+
+
 # v3.5.2
 
 ## New features

diff --git a/CHANGELOG_CN.MD b/CHANGELOG_CN.MD
@@ -1,3 +1,84 @@
+# v3.5.3
+
+## 新特性
+
+- Pika 支持 ACL[#2013](https://github.com/OpenAtomFoundation/pika/pull/2013) @[lqxhub](https://github.com/lqxhub)
+
+- 在 Codis dashboard 协程 panic 时自动恢复服务[#2349](https://github.com/OpenAtomFoundation/pika/pull/2349)@[chengyu-l](https://github.com/chengyu-l)
+
+- 在全量复制的过程中，pika 服务的从节点不接收读流量请求 [#2197](https://github.com/OpenAtomFoundation/pika/pull/2197) @[tedli](https://github.com/tedli)
+
+- Pika cache 新增 bimap数据类型[#2197](https://github.com/OpenAtomFoundation/pika/pull/2197) @[chejinge](https://github.com/chejinge)
+
+- 删除 Sharing 模式残留的 Slot，Pika 下只有 DB，一个 Pika 下有多个 DB[#2251](https://github.com/OpenAtomFoundation/pika/pull/2251) @[Mixficsol](https://github.com/Mixficsol)
+
+- Pika exporter 暴露 cache 相关的数据采集指标[#2318](https://github.com/OpenAtomFoundation/pika/pull/2318) @[dingxiaoshuai](https://github.com/dingxiaoshuai123)
+
+- Pika 支持快慢命令分离[#2162](https://github.com/OpenAtomFoundation/pika/pull/2162) @[dingxiaoshuai](https://github.com/dingxiaoshuai123)
+
+- pika 执行完成 Bgsave后, 保留 unix timepoint[#2167](https://github.com/OpenAtomFoundation/pika/pull/2167) @[hero-heng](https://github.com/hero-heng)
+
+- Pika 支持动态配置 disable_auto_compations 参数[#2257](https://github.com/OpenAtomFoundation/pika/pull/2257) @[hero-heng](https://github.com/hero-heng)
+
+- Pika 支持 Redis Stream[#1955](https://github.com/OpenAtomFoundation/pika/pull/1955) @[KKorpse](https://github.com/KKorpse)
+
+- Pika 支持大 key 分析工具[#2195](https://github.com/OpenAtomFoundation/pika/pull/2195) @[sjcsjc123](https://github.com/sjcsjc123)
+
+- Pika 支持动态调整 Pika cache 参数[#2197](https://github.com/OpenAtomFoundation/pika/pull/2197) @[chejinge](https://github.com/chejinge)
+
+- 更新 Pika benchmark 工具支持更多的接口压测[#2222](https://github.com/OpenAtomFoundation/pika/pull/2222)@[wangshao1](https://github.com/wangshao1)
+
+- Pika Operator 支持 pika 集群自动扩容[#2121](https://github.com/OpenAtomFoundation/pika/pull/2121)@[machinly](https://github.com/machinly/)
+
+- 添加 CompactRange 命令支持对一定范围内的 key 进行 compact[#2163](https://github.com/OpenAtomFoundation/pika/pull/2163)@[u6th9d](https://github.com/u6th9d)
+
+- 提升 Compaction 速度减少 Compaction 耗时[#2172](https://github.com/OpenAtomFoundation/pika/pull/2172)@[u6th9d](https://github.com/u6th9d)
+
+- 升级 RocksDB 版本到 v8.7.3[#2157](https://github.com/OpenAtomFoundation/pika/pull/2157)@[JasirVoriya](https://github.com/JasirVoriya)
+
+- Pika 分布式集群 Codis proxy 新增可观测指标[#2199](https://github.com/OpenAtomFoundation/pika/pull/2199)@[dingxiaoshuai](https://github.com/dingxiaoshuai123)
+
+- Pika 分布式集群支持自动 failover[#2386](https://github.com/OpenAtomFoundation/pika/pull/2386)@[chengyu-l](https://github.com/chengyu-l)
+
+## bugfix
+
+- 修复 Pika 有从节点进行全量复制期间会误删除 dump 文件的问题[#2377](https://github.com/OpenAtomFoundation/pika/pull/2377)@[wangshao1](https://github.com/wangshao1)
+
+- 修复主从复制过程中, slave 节点收到 master 异常回包后的处理逻辑[#2319](https://github.com/OpenAtomFoundation/pika/pull/2319)@[wangshao1](https://github.com/wangshao1)
+
+- 在 Pika 执行 shutdown 命令时调用 disable compaction, 提升进程退出速度 [#2345](https://github.com/OpenAtomFoundation/pika/pull/2345) @[panlei-coder](https://github.com/panlei-coder)
+
+- 修复 Codis-dashboard Redis Memory 值不准确的问题[#2337](https://github.com/OpenAtomFoundation/pika/pull/2337) @[Mixficsol](https://github.com/Mixficsol)
+
+- INFO 命令耗时优化，降低查磁盘频率 [#2197](https://github.com/OpenAtomFoundation/pika/pull/2197) @[chejinge](https://github.com/chejinge)
+
+- 修复 Rsync 删除临时文件路径不对，删除失败，导致rocksdb打开失败的问题[#2186](https://github.com/OpenAtomFoundation/pika/pull/2186)@[wangshao1](https://github.com/wangshao1)
+
+- 修复 Compact ,Bgsave ,Info keyspace 命令未指定db名称，导致部分命令 coredump 的问题[#2194](https://github.com/OpenAtomFoundation/pika/pull/2194)@[u6th9d](https://github.com/u6th9d)
+
+- Codis dashboard 用 info replication 替代 info 命令查寻 master ip 降低对 Pika 的性能影响 [#2198](https://github.com/OpenAtomFoundation/pika/pull/2198) @[chenbt-hz](https://github.com/chenbt-hz)
+
+- 修复 Pika cache 使用边缘case，解决部分场景下 cache 和 DB 数据不一致的问题[#2225](https://github.com/OpenAtomFoundation/pika/pull/2225) @[chejinge](https://github.com/chejinge)
+
+- 修复当 dump 文件夹为空时，会启动报错 Segmentation fault 的问题[#2265](https://github.com/OpenAtomFoundation/pika/pull/2265) @[chenbt-hz](https://github.com/chenbt-hz)
+
+- 修复因为flag计算错误,导致的部分命令缓存没有生效问题[#2217](https://github.com/OpenAtomFoundation/pika/pull/2217) @[lqxhub](https://github.com/lqxhub)
+
+- 修复主从复制模式下，主实例 flushdb 后，从实例因为死锁导致的不能访问的问题[#2249](https://github.com/OpenAtomFoundation/pika/pull/2249)@[ForestLH](https://github.com/ForestLH)
+
+- 修复部分命令未对 RocksDB 的返回值进行判断的问题[#2187](https://github.com/OpenAtomFoundation/pika/pull/2187)@[callme-taota](https://github.com/callme-taota)
+
+- 规范函数的返回值及初始值[#2176](https://github.com/OpenAtomFoundation/pika/pull/2176)@[Mixficsol](https://github.com/Mixficsol)
+
+- 修复网络监控指标统计不准确的问题[#2234](https://github.com/OpenAtomFoundation/pika/pull/2234)@[chengyu-l](https://github.com/chengyu-l)
+
+- 修复配置文件加载部分参数异常的问题[#2218](https://github.com/OpenAtomFoundation/pika/pull/2218)@[jettcc](https://github.com/jettcc)
+
+- 修复 Codis dashboard cpu 100% 的问题[#2393](https://github.com/OpenAtomFoundation/pika/pull/2393)@[chengyu-l](https://github.com/chengyu-l)
+
+- 修复 Codis fe pika 主从角色显示异常的问题[#2387](https://github.com/OpenAtomFoundation/pika/pull/2387)@[chengyu-l](https://github.com/chengyu-l)
+
+
 # v3.5.2
 
 ## 新特性

diff --git a/codis/config/dashboard.toml b/codis/config/dashboard.toml
@@ -33,9 +33,10 @@ migration_async_numkeys = 500
 migration_timeout = "30s"
 
 # Set configs for redis sentinel.
-sentinel_check_server_state_interval = "5s"
-sentinel_check_master_failover_interval = "1s"
-sentinel_master_dead_check_times = 5
+sentinel_check_server_state_interval = "10s"
+sentinel_check_master_failover_interval = "2s"
+sentinel_master_dead_check_times = 10
+sentinel_check_offline_server_interval = "2s"
 sentinel_client_timeout = "10s"
 sentinel_quorum = 2
 sentinel_parallel_syncs = 1

diff --git a/codis/pkg/models/action.go b/codis/pkg/models/action.go
@@ -11,4 +11,7 @@ const (
  ActionMigrating = "migrating"
  ActionFinished = "finished"
  ActionSyncing = "syncing"
+ ActionSynced = "synced"
+
+ ActionSyncedFailed = "synced_failed"
 )
diff --git a/codis/pkg/models/group.go b/codis/pkg/models/group.go
@@ -25,6 +25,38 @@ func (g *Group) GetServersMap() map[string]*GroupServer {
  return results
 }
 
+// SelectNewMaster choose a new master node in the group
+func (g *Group) SelectNewMaster() (string, int) {
+ var newMasterServer *GroupServer
+ var newMasterIndex = -1
+
+ for index, server := range g.Servers {
+ if index == 0 || server.State != GroupServerStateNormal {
+ continue
+ }
+
+ if newMasterServer == nil {
+ newMasterServer = server
+ newMasterIndex = index
+ } else if server.DbBinlogFileNum > newMasterServer.DbBinlogFileNum {
+ // Select the slave node with the latest offset as the master node
+ newMasterServer = server
+ newMasterIndex = index
+ } else if server.DbBinlogFileNum == newMasterServer.DbBinlogFileNum {
+ if server.DbBinlogOffset > newMasterServer.DbBinlogOffset {
+ newMasterServer = server
+ newMasterIndex = index
+ }
+ }
+ }
+
+ if newMasterServer == nil {
+ return "", newMasterIndex
+ }
+
+ return newMasterServer.Addr, newMasterIndex
+}
+
 type GroupServerState int8
 
 const (
@@ -33,6 +65,13 @@ const (
  GroupServerStateOffline
 )
 
+type GroupServerRole string
+
+const (
+ RoleMaster GroupServerRole = "master"
+ RoleSlave GroupServerRole = "slave"
+)
+
 type GroupServer struct {
  Addr string `json:"server"`
  DataCenter string `json:"datacenter"`
@@ -43,9 +82,11 @@ type GroupServer struct {
  } `json:"action"`
 
  // master or slave
- Role string `json:"role"`
+ Role GroupServerRole `json:"role"`
  // If it is a master node, take the master_repl_offset field, otherwise take the slave_repl_offset field
- ReplyOffset int `json:"reply_offset"`
+ DbBinlogFileNum uint64 `json:"binlog_file_num"` // db0
+ DbBinlogOffset uint64 `json:"binlog_offset"` // db0
+
  // Monitoring status, 0 normal, 1 subjective offline, 2 actual offline
  // If marked as 2 , no service is provided
  State GroupServerState `json:"state"`

diff --git a/codis/pkg/topom/config.go b/codis/pkg/topom/config.go
@@ -50,9 +50,10 @@ migration_async_numkeys = 500
 migration_timeout = "30s"
 
 # Set configs for redis sentinel.
-sentinel_check_server_state_interval = "5s"
-sentinel_check_master_failover_interval = "1s"
-sentinel_master_dead_check_times = 5
+sentinel_check_server_state_interval = "10s"
+sentinel_check_master_failover_interval = "2s"
+sentinel_master_dead_check_times = 10
+sentinel_check_offline_server_interval = "2s"
 sentinel_client_timeout = "10s"
 sentinel_quorum = 2
 sentinel_parallel_syncs = 1
@@ -86,6 +87,7 @@ type Config struct {
  SentinelCheckServerStateInterval timesize.Duration `toml:"sentinel_check_server_state_interval" json:"sentinel_client_timeout"`
  SentinelCheckMasterFailoverInterval timesize.Duration `toml:"sentinel_check_master_failover_interval" json:"sentinel_check_master_failover_interval"`
  SentinelMasterDeadCheckTimes int8 `toml:"sentinel_master_dead_check_times" json:"sentinel_master_dead_check_times"`
+ SentinelCheckOfflineServerInterval timesize.Duration `toml:"sentinel_check_offline_server_interval" json:"sentinel_check_offline_server_interval"`
  SentinelClientTimeout timesize.Duration `toml:"sentinel_client_timeout" json:"sentinel_client_timeout"`
  SentinelQuorum int `toml:"sentinel_quorum" json:"sentinel_quorum"`
  SentinelParallelSyncs int `toml:"sentinel_parallel_syncs" json:"sentinel_parallel_syncs"`

diff --git a/codis/pkg/topom/context.go b/codis/pkg/topom/context.go
@@ -40,7 +40,7 @@ func (ctx *context) getSlotMapping(sid int) (*models.SlotMapping, error) {
 }
 
 func (ctx *context) getSlotMappingsByGroupId(gid int) []*models.SlotMapping {
- var slots = []*models.SlotMapping{}
+ var slots []*models.SlotMapping
  for _, m := range ctx.slots {
  if m.GroupId == gid || m.Action.TargetId == gid {
  slots = append(slots, m)

diff --git a/codis/pkg/topom/topom.go b/codis/pkg/topom/topom.go
@@ -210,12 +210,12 @@ func (s *Topom) Start(routines bool) error {
  }
  }, nil, true, 0)
 
- // Check the status of the pre-offline master every 1 second
+ // Check the status of the pre-offline master every 2 second
  // to determine whether to automatically switch master and slave
  gxruntime.GoUnterminated(func() {
  for !s.IsClosed() {
  if s.IsOnline() {
- w, _ := s.CheckPreOffineMastersState(5 * time.Second)
+ w, _ := s.CheckPreOfflineMastersState(5 * time.Second)
  if w != nil {
  w.Wait()
  }
@@ -224,6 +224,20 @@ func (s *Topom) Start(routines bool) error {
  }
  }, nil, true, 0)
 
+ // Check the status of the offline master and slave every 30 second
+ // to determine whether to automatically recover to right master-slave replication relationship
+ gxruntime.GoUnterminated(func() {
+ for !s.IsClosed() {
+ if s.IsOnline() {
+ w, _ := s.CheckOfflineMastersAndSlavesState(5 * time.Second)
+ if w != nil {
+ w.Wait()
+ }
+ }
+ time.Sleep(s.Config().SentinelCheckOfflineServerInterval.Duration())
+ }
+ }, nil, true, 0)
+
  gxruntime.GoUnterminated(func() {
  for !s.IsClosed() {
  if s.IsOnline() {