myrocks basic usage

[jd-zhang]

*Issue migrated from trac ticket # 823 www.kunlunbase.com *

component: Testing | priority: major

2022-06-15 09:57:02: zhaowei@zettadb.com created the issue

---

0. Read this first: https://www.percona.com/blog/2018/02/01/myrocks-engine-things-know-start/
1. create a cluster with several shards, manually alter each table to use myrocks: alter table t1 engine='myrocks';
2. DDL & DML operations, using tpcc and sysbench
3. record QPS/TPS and latency and data compression ratio
4. load massive(500GB) data using SQL stmts
5. check binlog replication work correctly
6. do cluster physical backup&restore
7. do scale out

[jd-zhang]

2022-06-15 10:18:39: zhaowei@zettadb.com edited the issue description

[jd-zhang]

2022-06-22 12:14:14: hegexin commented

---

测试环境： 1主两从 （32c , 128g mem)

1， 压缩率
sysbench 1.0 , 10张表， 50000万记录/表， innodb 12247104kB, rocksdb 9019808KB , 压缩率73.6%

2， tpcc 测试。

100并发： oltp_write_only.lua ， tps: 283.40 per sec.
200并发： oltp_write_only.lua ， tps: 8092 per sec.
300并发： oltp_write_only.lua ， tps: 11427.27 per sec

增加线程， fullsync 超时错误 （等待解决。。。）

FATAL: `thread_run' function failed: ...unlun/tmp/sysbench/output/share/sysbench/oltp_common.lua:409: SQL error, errno # 9000, state'HY000': Fullsync in thread 10279 timed out running command COM_STMT_EXECUTE: 'COMMIT', got 0 ACKs.

3，复制延迟

100并发 <10S
200并发 <30s