original status of being enabled or disabled after it is migrated to any DM-worker. – Move the storage path of relay log to the DM-worker configuration file. User document • Optimize the processing of collation document • Optimize transfer source in DM to support running replication tasks smoothly When the DM-worker nodes have an unbalanced load, the transfer source command can be used to manually transfer the configuration enabled: true tikv-client.max-batch-wait-time: 2000000 TiKV parameter configuration storage.scheduler-worker-pool-size: 5 raftstore.store-pool-size: 3 raftstore.apply-pool-size: 3 rocksdb.max-background-jobs:

stop-relay 命令动态开启和关闭 relay log 的功能。 – relay log 的开启状态与 source 绑定，source 迁移到任意 DM-worker 均保持原有开启或关闭状态。 – relay log 的存放路径移至 DM-worker 配置文件。 用户文档 • 在 DM 中优化排序规则的处理方式 增加 collation_compatible 开关，支持 strict 和 loose（默认）两种模式： 用 strict 模式。但如果下游不支 持上游缺省的 collation，同步可能会报错。 用户文档 • 在 DM 中优化 transfer source，支持平滑执行同步任务 当 DM-worker 所在各节点负载不均衡时，transfer source 命令可用于手动将某 source 配置迁移到其他 节点。优化后的 transfer source 简化了用户操作步骤，不再要求先暂停所有关联 prepared-plan-cache.enabled: true tikv-client.max-batch-wait-time: 2000000 TiKV 参数配置 storage.scheduler-worker-pool-size: 5 raftstore.store-pool-size: 3 raftstore.apply-pool-size: 3 rocksdb.max-background-jobs:

point that stores Avro schema. 78 Configuration file Configuration Change type Description DM worker �→ in the dmctl �→ �→ start �→ - �→ relay �→ com- mand DeletedThis pa- ram- e- ter is not rec- om- source con- fig- u- ra- tion file DeletedReplaced by the same con- fig- u- ra- tion item in the worker con- fig- u- ra- tion file. 79 Configuration file Configuration Change type Description DM is- uppercase table cannot be replicated when case- �→ sensitive: true is not set #5255 • Fix the DM worker panic issue that occurs when the primary key is not first in the index returned by the SHOW CREATE

able is set to 0, no limit is ap- plied. 64 Variable name Change type Description tidb_ttl_delete_worker_count �→ Newly added This vari- able is used to set the maxi- mum concur- rency of TTL jobs on each the LIMIT value of each SELECT state- ment used to scan expired data in a TTL job. tidb_ttl_scan_worker_count �→ Newly added This vari- able is used to set the maxi- mum concur- rency of TTL scan jobs 88 • Fix the issue that data is replicated for multiple times when a new DM worker is scheduled before the existing worker exits #7658 @GMHDBJD • Fix the issue that DM precheck is not passed when the

#49318 @bufferflies • Support dynamically adjusting the value of readpool.unified.max-tasks-per �→ -worker, which can calculate the number of running tasks separately based on priority #16026 @glorv • Support partial_worker:{wall_time:60.660079ms, �→ concurrency:5, task_num:293, tot_wait:262.536669ms, tot_exec �→ :40.171833ms, tot_time:302.827753ms, max:60.636886ms, p95:60.636886ms �→ }, final_worker:{wall_time:60 Look at the TTL Scan Worker Time By Phase and TTL Delete Worker Time By �→ Phase panels. If the scan worker is in the dispatch phase for a large percentage of time and the delete worker is rarely in the idle

excep- tions occur. TiCDC encoding �→ - �→ worker �→ -num Newly added Controls the number of encoding and decoding workers in the redo module. TiCDC flush- �→ worker �→ -num Newly added Controls the number partial_worker:{wall_time:60.660079ms, �→ concurrency:5, task_num:293, tot_wait:262.536669ms, tot_exec �→ :40.171833ms, tot_time:302.827753ms, max:60.636886ms, p95:60.636886ms �→ }, final_worker:{wall_time:60 Look at the TTL Scan Worker Time By Phase and TTL Delete Worker Time By �→ Phase panels. If the scan worker is in the dispatch phase for a large percentage of time and the delete worker is rarely in the idle

of a row on a partitioned table with a global index #42312 @L-maple • Fix the issue that the Scan Worker Time By Phase chart in the TTL monitoring panel does not display data #42515 @lcwangchao • Fix the partial_worker:{wall_time:60.660079ms, �→ concurrency:5, task_num:293, tot_wait:262.536669ms, tot_exec �→ :40.171833ms, tot_time:302.827753ms, max:60.636886ms, p95:60.636886ms �→ }, final_worker:{wall_time:60 Look at the TTL Scan Worker Time By Phase and TTL Delete Worker Time By �→ Phase panels. If the scan worker is in the dispatch phase for a large percentage of time and the delete worker is rarely in the idle

Endpoint。 DM dmctl �→ �→ start �→ - �→ relay �→ 命 令中 的 worker 参数 删除 不推荐 使用的 方式， 将通过 更为简 单的实 现替代。 DM source 配置 中的 relay �→ - �→ dir �→ 删除 由 worker 配置文 件中的 同名配 置项替 代。 DM task 配 置中 的 is- �→ sharding �→ 修复在未设置 case-sensitive: true 时无法同步大写表的问题 #5255 * 修复了在 SHOW CREATE TABLE 语句返回的索引中，主键没有排在第一位导致的 DM worker panic 的问题 #5159 * 修复了当开启 GTID 模式或者任务自动恢复时，可能出现一段时间 CPU 占用高并打印大量日志 的问题 #5063 * 修复 DM Web UI offline prepared-plan-cache.enabled: true tikv-client.max-batch-wait-time: 2000000 TiKV 参数配置 64 storage.scheduler-worker-pool-size: 5 raftstore.store-pool-size: 3 raftstore.apply-pool-size: 3 rocksdb.max-background-jobs:

#42659 @hawkingrei – 修复在包含全局索引的分区表上更新分区键数据时报 Duplicate Key 错误的问题 #42312 @L-maple – 修复 TTL 监控面板中 Scan Worker Time By Phase 图表不显示数据的问题 #42515 @lcwangchao – 修复在包含全局索引的分区表上进行某些查询时返回错误结果的问题 #41991 #42065 @L-maple partial_worker:{wall_time:60.660079ms, concurrency:5, task_num:293, tot_wait �→ :262.536669ms, tot_exec:40.171833ms, tot_time:302.827753ms, max:60.636886ms, p95 �→ :60.636886ms}, final_worker:{wall_time:60 任务的瓶颈在扫描还是删除？ 观察面板中 TTL Scan Worker Time By Phase 与 TTL Delete Worker Time By Phase 监控项。如果 scan worker 处于 dispatch 状态的时间有很大占比，且 delete worker 很少处于 idle 状态，那么说明 scan worker 在等待 delete worker 完成删除工作，如果此时集群资源仍然较为宽松，可以考虑提高

秒允许 DELETE 语句执 行的最 大次数。 当此变 量设置 为 0 时， 则表示 不做限 制。 49 变量名 修改类型 描述 tidb_ttl_ �→ delete �→ _ �→ worker �→ _count 新增 这个变 量用于 设置每 个 TiDB 节点上 TTL 删除 任务的 最大并 发数。 tidb_ttl_ �→ job_ �→ enable 新增 这个变 量用于 batch_ �→ size 新增 这个变 量用于 设置 TTL 任务中 用来扫 描过期 数据的 每个 SELECT 语句的 LIMIT 的 值。 tidb_ttl_ �→ scan_ �→ worker �→ _count 新增 这个变 量用于 设置每 个 TiDB 节点 TTL 扫描任 务的最 大并发 数。 52 变量名 修改类型 描述 validate_ �→ password �→ Migration (DM) * 修复在上游开启 GTID mode 且无数据时，无法启动 all mode 任务的问题 #7037 @liumengya94 * 修复当 DM worker 即将退出时新 worker 调度过快导致数据被重复同步的问题 #7658 @GMHDBJD * 修复上游数据库使用正则匹配授权时 DM 前置检查不通过的错误 #7645 @lance6716 – TiDB