Apache HBaseの現在 - 火山と呼ばれたHBaseは今どうなっているのか
0 likes2,380 views
The document discusses several key topics in Apache HBase: 1. Procedure version 2 introduces a new framework for running operations like create/drop table and region assignment as procedures with distinct phases. 2. Assignment Manager version 2 uses procedures and improves region assignment and load balancing. 3. Backup/restore now supports HDFS, S3, ADLS and WASB. Snapshots can also be used for backup. 4. Compacting memstore allows in-memory flushing and compaction to improve performance through pipelining.
![Backup/Restore
•
•
• hbase backup create <type> <backup_path> [options]
• hbase restore <backup_path> <backup_id> [options]
• HDFS S3, ADLS, WASB
•
• hbase snapshot
• Write Ahead Log (WAL)](https://image.slidesharecdn.com/hjc2019-190315015242/85/Apache-HBase-HBase-16-320.jpg)
Ad
More Related Content
What's hot (20)
Similar to Apache HBaseの現在 - 火山と呼ばれたHBaseは今どうなっているのか (20)
Ad
More from Toshihiro Suzuki (10)
Ad
Recently uploaded (20)
Apache HBaseの現在 - 火山と呼ばれたHBaseは今どうなっているのか
- 1. Apache HBase HBase Cloudera Hadoop / Spark Conference Japan 2019
- 2. ( ) • Apache HBase Committer • Cloudera • Sr. Software Engineer, Breakfix • • • ( HBase/Phoenix) • HBase • Twitter: @brfrn169
- 3. • HBase • HBase • HBase
- 4. HBase
- 6. • HBase 0.98 • HBase 1.4.9 • • HBase 1.5.0 • HBase 1 • HBase 2 HBase 2.1.x HBase 2.2.0 • HBase 2
- 7. • CDH • CDH 5.8+: HBase 1.2.0 (+ bugfixes and backports) • CDH 6.0: HBase 2.0.1 (+ bugfixes and backports) • CDH 6.1: HBase 2.1.1 (+ bugfixes and backports) • HDP • HDP 2.x: HBase 1.1.2 (+ bugfixes and backports) • HDP 3.x: HBase 2.0.2 (+ bugfixes and backports)
- 8. HBase
- 9. HBase • HBase 2.x • • Procedure version 2 • Assignment Manager version 2 • • Backup/Restore • • Compacting Memstore • • Serial Replication
- 10. Procedure version 2 • Master (create/drop table region assign split ) • Master Procedure
- 11. Procedure version 2 • ) CreateTableProcedure PRE_OPERATION WRITE_FS_LAYOUT ADD_TO_META ASSIGN_REGIONSUPDATE_DESC_CACHEPOST_OPERATION Start End
- 12. Procedure version 2 • ) CreateTableProcedure PRE_OPERATION WRITE_FS_LAYOUT ADD_TO_META ASSIGN_REGIONSUPDATE_DESC_CACHEPOST_OPERATION Start End
- 13. Procedure version 2 • ) CreateTableProcedure PRE_OPERATION WRITE_FS_LAYOUT ADD_TO_META ASSIGN_REGIONSUPDATE_DESC_CACHEPOST_OPERATION Start End
- 14. Procedure version 2 • ) CreateTableProcedure PRE_OPERATION WRITE_FS_LAYOUT ADD_TO_META ASSIGN_REGIONSUPDATE_DESC_CACHEPOST_OPERATION Start End Procedure ASSIGN_REGIONS Region Procedure
- 15. Assignment Manager version 2 • Region • Region • HBCK • Region Assignment Manager version 2 • Procedure version 2 • Region Zookeeper • • • Region • Region • Master
- 16. Backup/Restore • • • hbase backup create <type> <backup_path> [options] • hbase restore <backup_path> <backup_id> [options] • HDFS S3, ADLS, WASB • • hbase snapshot • Write Ahead Log (WAL)
- 17. Compacting Memstore • Compacting Memstore • in-memory flush • • in-memory compaction • • Flush • Compaction
- 18. Compacting Memstore • Default Memstore ( ) Active HDFS
- 19. Compacting Memstore • Default Memstore ( ) ActiveWrite HDFS
- 20. Compacting Memstore • Default Memstore ( ) ActiveWrite Snapshot HDFS Active
- 21. Compacting Memstore • Default Memstore ( ) ActiveWrite Snapshot HDFS Flush HFile Active
- 22. Compacting Memstore • Default Memstore ( ) Active HDFS HFile HFile
- 23. Compacting Memstore • Default Memstore ( ) Active HDFS HFile HFile Compaction HFile
- 24. Compacting Memstore • Compacting Memstore Active HDFS
- 25. Compacting Memstore • Compacting Memstore ActiveWrite HDFS
- 26. Compacting Memstore • Compacting Memstore ActiveWrite HDFS Pipeline #1 Active in-memory flush
- 27. Compacting Memstore • Compacting Memstore ActiveWrite HDFS Pipeline #1 Pipeline #2 Pipeline #3
- 28. Compacting Memstore • Compacting Memstore ActiveWrite HDFS Pipeline #1 Pipeline #2 Pipeline #3 Pipeline in-memory compaction
- 29. Compacting Memstore • Compacting Memstore ActiveWrite HDFS Pipeline #1 Pipeline in-memory compaction
- 30. Compacting Memstore • Compacting Memstore ActiveWrite HDFS Pipeline #1 Pipeline #2
- 31. Compacting Memstore • Compacting Memstore ActiveWrite HDFS Pipeline #1 Pipeline #2 Flush HFile
- 32. Serial Replication • HBase Replication • • • Push • ( ) ( ) Push • MySQL Pull
- 33. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously
- 34. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously Tail the WALs
- 35. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously Tail the WALs
- 36. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1 Tail the WALs
- 37. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1
- 38. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1
- 39. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 2 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1
- 40. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 2 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1 2
- 41. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1 2
- 42. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1 2
- 43. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 3 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1 2
- 44. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 3 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1 2 3
- 45. Serial Replication • HBase Replication RegionServer WAL1 WAL2 1 Queue 2 3 4 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push asynchronously 1 2 3
- 46. Serial Replication • HBase Replication • RegionServer Region move Push
- 47. Serial Replication • HBase Replication RegionServer 1Queue ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource
- 48. Serial Replication • HBase Replication RegionServer 1 1 Queue ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource
- 49. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource
- 50. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource
- 51. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource Move the Region
- 52. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource 4 Move the Region
- 53. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push RegionServer 2Queue ReplicationSource 4 Push
- 54. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push RegionServer 2Queue ReplicationSource 4 Push
- 55. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push RegionServer 2Queue ReplicationSource 4 Push 1
- 56. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 Push 1
- 57. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 Push
- 58. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 Push
- 59. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 Push 2
- 60. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 2
- 61. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push
- 62. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push
- 63. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 4
- 64. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 4 4
- 65. Serial Replication • HBase Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 4 Inconsistent State!
- 66. Serial Replication • Serial Replication • Serial Replication
- 70. Serial Replication • HBase RegionServer Region WAL Client Put MemStore HDFS 1 Assign Sequence ID to the data (Cell) before writing WAL
- 74. Serial Replication • Sequence ID • Region (Cell) • Multi Version Concurrency Control (MVCC) • Serial Replication
- 75. Serial Replication RegionServer 1Queue ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource Zookeeper hbase:meta
- 76. Serial Replication RegionServer 1 1 Queue ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource Zookeeper hbase:meta
- 77. Serial Replication RegionServer 1 1 Queue 2 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource Zookeeper hbase:meta
- 78. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource Zookeeper hbase:meta
- 79. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource Zookeeper hbase:meta Move the Region
- 80. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource Zookeeper hbase:meta The sequence of open sequence numbers for the region 3
- 81. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource Cluster 1 RegionServer ReplicationSink Cluster 2 HTable RegionServer RegionServer 2Queue ReplicationSource 4 Zookeeper hbase:meta 3
- 82. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 3
- 83. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 3
- 84. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push RegionServer 2Queue ReplicationSource 4 Push 1 Zookeeper hbase:meta 3
- 85. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 Push 1 Zookeeper hbase:meta 3
- 86. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 31 The last pushed Sequence ID
- 87. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 31
- 88. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 31
- 89. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 RegionServer 2Queue ReplicationSource 4 Push 2 Zookeeper hbase:meta 31
- 90. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 2 Zookeeper hbase:meta 31
- 91. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 31 The last pushed Sequence ID 2
- 92. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 312
- 93. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 312
- 94. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 312 Wait
- 95. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 312
- 96. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper hbase:meta 312
- 97. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 3 Zookeeper hbase:meta 312
- 98. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 3 Zookeeper 3 hbase:meta 312
- 99. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper 3 hbase:meta 31 The last pushed Sequence ID 23
- 100. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper 3 hbase:meta 3123
- 101. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper 3 hbase:meta 3123
- 102. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 PushZookeeper 3 hbase:meta 3123 Go
- 103. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 4 Zookeeper 3 hbase:meta 3123
- 104. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 4 4 Zookeeper 3 hbase:meta 3123
- 105. Serial Replication RegionServer 1 1 Queue 2 3 ReplicationSource RegionServer ReplicationSink HTable RegionServer Push 1 2 RegionServer 2Queue ReplicationSource 4 Push 4 Zookeeper 3 hbase:meta 3123
- 106. • Procedure version 2 / Assignment Manager version 2 • • Backup/Restore • Compacting Memstore • Serial Replication Replication
- 107. HBase
- 108. HBase • Evolving HBase in the Cloud • HBase • HBase on Persistent Memory • HBase Persistent Memory • Synchronous Replication • • •
- 109. HBase • Evolving HBase in the Cloud • HBase • HBase on Persistent Memory • HBase Persistent Memory • Synchronous Replication • • •
- 110. Evolving HBase in the Cloud • HBASE-20951 Ratis LogService backed WALs • IaaS (Amazon EC2, Google Compute Engine, Microsoft Azure Compute) HBase • IaaS HBase
- 111. Evolving HBase in the Cloud • IaaS • Amazon EC2 • AWS • HDFS • DataNode • AWS
- 112. Evolving HBase in the Cloud • Amazon EBS (Elastic Block Store) Google Persistent Storage ( ) • Amazon EBS (Elastic Block Store) Google Persistent Storage • Amazon S3 Google Cloud Storage • • Amazon EBS Google Persistent Storage
- 113. Evolving HBase in the Cloud • HBase HFile WAL HDFS • HFile • WAL short-lived, sub-second durability requirements HDFS HFile HFile HFile WAL RegionServerPuts Memstore Flush
- 114. Evolving HBase in the Cloud • HFile (S3 with S3Guard ) • WAL • WAL • sub-second durability requirements • WAL • traversable queue (FIFO) • constant-time append complexity • linear-time traversal • sub-linear seek to an arbitrary offset
- 115. Evolving HBase in the Cloud • Apache Ratis • Apache Software Foundation • RAFT Java • Apache Hadoop Ozone • Ratis Kafka DistributedLog • HBase WAL • • Ratis • WAL Ratis
- 116. Evolving HBase in the Cloud • Ratis WAL Ratis LogService Ratis • WAL HBase • 2 1. Ratis LogService (RATIS-271) 2. HBase WAL (HBASE-20952) • HDFS HDFS WAL 1 • Ratis LogService Kafka DistributedLog
- 117. Evolving HBase in the Cloud • RegionServer1 ReginoServer2 New WAL API Ratis LogService Amazon S3/Google Cloud Storage ReginoServer3 Flush Memstore WAL Storage WAL Storage WAL Storage Puts HFile HFile HFile RAFT