pgbench --help pgbench is a benchmarking tool for PostgreSQL. Usage:  pgbench [OPTION]... [DBNAME] Initialization options:  -i, --initialize         invokes initialization mode 初始化模式  -F, --fillfactor=NUM     set fill factor  -n, --no-vacuum          do not run VACUUM after initialization  -q, --quiet              quiet logging (one message each 5 seconds)  -s, --scale=NUM          scaling factor  --foreign-keys           create foreign key constraints between tables      --index-tablespace=TABLESPACE                               create indexes in the specified tablespace      --tablespace=TABLESPACE  create tables in the specified tablespace      --unlogged-tables        create tables as unlogged tables        Options to select what to run:      -b, --builtin=NAME[@W]   add builtin script NAME weighted at W (default: 1)                               (use "-b list" to list available scripts)      -f, --file=FILENAME[@W]  add script FILENAME weighted at W (default: 1)  测试脚本位置（可以自定义）      -N, --skip-some-updates  skip updates of pgbench_tellers and pgbench_branches                               (same as "-b simple-update")      -S, --select-only        perform SELECT-only transactions                               (same as "-b select-only")        Benchmarking options:      -c, --client=NUM         number of concurrent database clients (default: 1)  模拟客户端个数，相当于并发量  -C, --connect            establish new connection for each transaction  设定为每一个事务建一个新连接     -D, --define=VARNAME=VALUE                               define variable for use by custom script      -j, --jobs=NUM           number of threads (default: 1)    启动线程数  -l, --log                write transaction times to log file      -L, --latency-limit=NUM  count transactions lasting more than NUM ms as late      -M, --protocol=simple|extended|prepared                               protocol for submitting queries (default: simple)      -n, --no-vacuum          do not run VACUUM before tests      -P, --progress=NUM       show thread progress report every NUM seconds    每隔数秒显示线程进度报告  -r, --report-latencies   report average latency per command    报告每个命令的平均延迟  -R, --rate=NUM           target rate in transactions per second      -s, --scale=NUM          report this scale factor in output      -t, --transactions=NUM   number of transactions each client runs (default: 10)      -T, --time=NUM           duration of benchmark test in seconds   测试时间，单位：秒   -v, --vacuum-all         vacuum all four standard tables before tests      --aggregate-interval=NUM aggregate data over NUM seconds      --progress-timestamp     use Unix epoch timestamps for progress      --sampling-rate=NUM      fraction of transactions to log (e.g., 0.01 for 1%)     Common options:  -d, --debug              print debugging output      -h, --host=HOSTNAME      database server host or socket directory      -p, --port=PORT          database server port number      -U, --username=USERNAME  connect as specified database user      -V, --version            output version information, then exit  -?, --help               show this help, then exit
复制

su - postgres # 使用postgres作为测试库pgbench -i -d postgres # 使用默认参数创造的测试表，各表的数据量：pgbench_accounts：100000pg_branches：1pg_tellers：10
复制

pgbench -i -F 100 -s 100 -d postgres
复制

pgbench -c 1 -T 20 -r -d postgres
复制

create table test(id int primary key, info text, crt_time timestamp);
复制

vi test.sql\set id random(1,100000000)insert into test (id,info,crt_time) values (:id, md5(random()::text), now()) on conflict (id) do update set info=excluded.info, crt_time=excluded.crt_time;
复制

pgbench -M prepared -r  -f ./test.sql -c 32 -j 32 -T 20 -d postgres
复制