Longda's Interesting World

Distributed Databases & AI Agent Engineering Practice

OceanBase-4.2.x Experience Notes (2)

OceanBase 4.2.x experience notes (part 2): under the 4.2.1 kernel and OBD 2.4.0 environment, recording the Alibaba Cloud ECS hardware selection and the three-replica OBServer, OBProxy/OCP/OMS deployment topology.

Summary

Test Outline

  1. OBD installation and deployment
  2. OBD inspection optimization
  3. tpcc testing
  4. sysbench testing
  5. tpch testing
  6. obdiag log collection
  7. Using OBD, scale up a single node into cluster mode
  8. Using OBD, vertically scale up the cluster (add machines within a single replica)
  9. Using OBD, horizontally scale in the cluster (remove replicas)
  10. Using OBD, horizontally scale out the cluster (add replicas)
  11. Using OBD, vertically scale down the cluster (remove machines within a single replica)
  12. failover test, when running sysbench, restart one server.
  13. ocp installation and deployment
  14. ocp takes over an OBD-deployed cluster
  15. ocp installs an OB cluster (single-node primary-standby mode)
  16. Using OBD, scale up a single node into cluster mode
  17. Using OBD, vertically scale up the primary cluster (add machines within a single replica)
  18. Using OBD, horizontally scale out the primary cluster (add replicas)
  19. Using OBD, horizontally scale in the primary cluster (remove replicas)
  20. Using OBD, vertically scale down the primary cluster (remove machines within a single replica)
  21. Using NFS, back up the primary cluster and restore the standby cluster
  22. Using NFS, back up the standby cluster; after a switchover, restore the original primary cluster
  23. oms installation and deployment
  24. tpch data synchronization to MySQL
  25. binlog service installation
  26. Under a sysbench test, configure the binlog service to synchronize to MySQL

Software Versions

  • Kernel version: 4.2.1_CE_BP1_HF1
  • OBD version: 2.4.0
  • OCP version: 4.2.1
  • OMS version: 4.2.0
  • obdiag diagnostic tool: 1.4.0
  • OBLogProxy (binlog service): v2.0.0

Hardware Configuration

The machine configuration running observer (since this is only a functional test, no higher-spec hardware was chosen)

  • aliyun ecs:
  • 4c/32g amd instance
  • 3 disks
  • system disk: essd 40g, IOPS limit 10,000
  • clog disk: essd 100g, IOPS limit 50,000
  • data disk: essd 400g, IOPS limit 50,000

Configuration of the other machines not running observer

  • aliyun ecs:
  • 4c/32g amd instance
  • 3 disks
  • system disk: essd 100g, IOPS limit 10,000

Deployment Topology

  • 1 machine: all clients (sysbench, tpch, tpcc) + obproxy + obd + obdiag
  • 3 observers
  • 1 for ocp, 1 for the ocp metadb running ob
  • 1 for oms