mxgate是gpcopy同步速度的2倍

Posted 盒马coding

tags:

篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了mxgate是gpcopy同步速度的2倍相关的知识，希望对你有一定的参考价值。

目录

目录

​

集群的架构

1、构造测试的数据

2、查看数据的大小和行数

3、创建表语句

4、查看创建后的表

5、编写mxgate同步脚本

6、编写gpcopy同步脚本

7、查看硬件资源

7、1 CPU和内存信息

7、2 网络信息 

8、常见问题总结

 8、1 slice bounds out of range

8、2 "gpcopy_temp_3" does not exist

8、3 Control connection pair should call `InitConnections` first

8、4 解决方式

参考资料MatrixDB - 时序数据库加载性能测试过程进入四维纵横yMatrix官网，体验超融合时序数据库MatrixDB。https://ymatrix.cn/blog/load-test-steps

集群的架构

序列数据库服务器segmentmaster访问端口
1Greenplum6.0.0sdw216 primary + 16 mirrorsdw35432
2sdw3
3sdw4
4matrixdb4.2sdw516 primary + 16 mirrorsdw75432
5sdw6
6sdw7

1、构造测试的数据

# 造10列100w设备的数据
time ./tsbs_generate_data --format="matrixdb" --use-case="massiveiot" --scale="1000000" --log-interval=60s --column-data-type=simple --point-type=int --columns-per-row=10 --timestamp-start="2016-01-01T00:00:00Z" --timestamp-end="2016-01-01T02:00:00Z" > /data/c1/cpu_10_100w.csv

# 造50列100w设备的数据
time ./tsbs_generate_data --format="matrixdb" --use-case="massiveiot" --scale="1000000" --log-interval=60s --column-data-type=simple --point-type=int --columns-per-row=50 --timestamp-start="2016-01-01T00:00:00Z" --timestamp-end="2016-01-01T02:00:00Z" > /data/c1/cpu_50_100w.csv

# 造100列100w设备的数据
time ./tsbs_generate_data --format="matrixdb" --use-case="massiveiot" --scale="1000000" --log-interval=60s --column-data-type=simple --point-type=int --columns-per-row=100 --timestamp-start="2016-01-01T00:00:00Z" --timestamp-end="2016-01-01T02:00:00Z" > /data/c1/cpu_100_100w.csv

2、查看数据的大小和行数

$ du -sh cpu_10_100w.csv
15G     cpu_10_100w.csv

$ du -sh cpu_50_100w.csv
63G     cpu_50_100w.csv

$ du -sh cpu_100_100w.csv
122G    cpu_100_100w.csv


$ wc -l cpu_10_100w.csv
120000000 cpu_10_100w.csv

$ wc -l cpu_50_100w.csv
120000000 cpu_50_100w.csv

$ wc -l cpu_100_100w.csv
120000000 cpu_100_100w.csv

3、创建表语句

# 创建创建表的function
create or replace function f(name, int, text) returns void as $$   
declare   
  res text := 'ts varchar(50),tag_id integer,';   
begin   
  for i in 1..$2 loop 
  res := res||'c'||i||' integer,';   
  end loop;   
  res := rtrim(res, ',');   
  if $3 = 'ao_col' then   
    res := 'create table '||$1||'('||res||') with(appendonly=true, compresstype=zstd, orientation=column) DISTRIBUTED BY (tag_id)';   
  elsif $3 = 'ao_row' then   
    res := 'create table '||$1||'('||res||') with(appendonly=true, blocksize=262144, orientation=row) DISTRIBUTED BY (tag_id)';   
  elsif $3 = 'heap_row' then   
    res := 'create table '||$1||'('||res||') with(appendonly=false) DISTRIBUTED BY (tag_id)';   
  else   
    raise notice 'use ao_col, ao_row, heap_row as $3';   
    return;   
  end if;   
  execute res;   
end;   
$$ language plpgsql;


# 创建表语句和加载语句

$ cat matrixdball--ao.sh
#bin/bash

col_arr=( 10 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_ao"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','ao_row');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_10_100w_ao FROM '/home/gpadmin/1c/cpu_10_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done


col_arr=( 50 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_ao"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','ao_row');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_50_100w_ao FROM '/home/gpadmin/1c/cpu_50_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done


col_arr=( 100 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_ao"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','ao_row');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_100_100w_ao FROM '/home/gpadmin/1c/cpu_100_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done



$ cat matrixdball--aoco.sh
#bin/bash

col_arr=( 10 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_aoco"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','ao_col');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_10_100w_aoco FROM '/home/gpadmin/1c/cpu_10_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done


col_arr=( 50 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_aoco"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','ao_col');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_50_100w_aoco FROM '/home/gpadmin/1c/cpu_50_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done


col_arr=( 100 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_aoco"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','ao_col');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_100_100w_aoco FROM '/home/gpadmin/1c/cpu_100_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done



$ cat matrixdball--heap.sh
#bin/bash

col_arr=( 10 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_heap"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','heap_row');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_10_100w_heap FROM '/home/gpadmin/1c/cpu_10_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done


col_arr=( 50 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_heap"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','heap_row');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_50_100w_heap FROM '/home/gpadmin/1c/cpu_50_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done


col_arr=( 100 )
device_arr=( 100 )

for col in "$col_arr[@]"
do
   for device in "$device_arr[@]"
   do
      echo "$col,$device heap"
          file_name="cpu_$col_$devicew"
          table_name="cpu_$col_$devicew_heap"

          echo "$col,$device heap_index"
          psql -d gpdb -c  "drop table if exists $table_name"
          psql -d gpdb -c  "select f('$table_name','$col','heap_row');"
          psql -d gpdb -c  "create index on $table_name (tag_id, ts);create index on $table_name (ts);"
          time psql -d gpdb -c  "COPY cpu_100_100w_heap FROM '/home/gpadmin/1c/cpu_100_100w.csv' WITH csv DELIMITER '|' NULL as 'null string'"
   done
done

4、查看创建后的表

gpdb=# \\dt*
                          List of relations
 Schema |       Name        | Type  |  Owner  |       Storage
--------+-------------------+-------+---------+----------------------
 public | cpu_100_100w_ao   | table | mxadmin | append only
 public | cpu_100_100w_aoco | table | mxadmin | append only columnar
 public | cpu_100_100w_heap | table | mxadmin | heap
 public | cpu_10_100w_ao    | table | mxadmin | append only
 public | cpu_10_100w_aoco  | table | mxadmin | append only columnar
 public | cpu_10_100w_heap  | table | mxadmin | heap
 public | cpu_50_100w_ao    | table | mxadmin | append only
 public | cpu_50_100w_aoco  | table | mxadmin | append only columnar
 public | cpu_50_100w_heap  | table | mxadmin | heap

5、编写mxgate同步脚本

以下脚本可以同步不同的表类型，heap、ao、aoco、分区表等

$  vim sync_mxgate.sh

#!bin/bash

mxgate --source transfer \\
       --src-host 192.168.***.** \\
       --src-port 5432 \\
       --src-db postgres \\
       --src-user gpadmin \\
       --src-password  gpadmin \\
       --src-schema public \\
       --src-table cpu_100_100w_heap \\
       --compress "lz4" \\
       --local-ip 192.168.***.*** \\
       --db-database postgres \\
       --db-master-port 5432 \\
       --db-user mxadmin \\
       --db-password mxadmin \\
       --target public.cpu_100_100w_heap \\
       --format csv \\
       --parallel 300 \\
       --interval 100  \\
       --stream-prepared 6 \\
       --time-format raw

6、编写gpcopy同步脚本

gpcopy --source-host 192.168.***.*** \\
        --source-port 5432 \\
        --source-user gpadmin \\
        --include-table postgres.public.cpu_100_100w_heap \\
        --dest-host 192.168.***.*** \\
        --dest-port 5432 \\
        --dest-user mxadmin \\
        --drop --exclude-table postgres.public.cpu_100_100w_heap \\
        --jobs 200 /
        --on-segment-threshold 90000

7、查看硬件资源

7、1 CPU和内存信息

 

7、2 网络信息 

8、常见问题总结

 8、1 slice bounds out of range

20211021:10:42:20 gpcopy:mxadmin:sdw7:026691-[INFO]:-Copy Timestamp = 20211021104220
20211021:10:42:20 gpcopy:mxadmin:sdw7:026691-[INFO]:-pg_dump (PostgreSQL) 12
20211021:10:42:20 gpcopy:mxadmin:sdw7:026691-[INFO]:-pg_dumpall (PostgreSQL) 12
20211021:10:42:21 gpcopy:mxadmin:sdw7:026691-[WARNING]:-pg_dump version is higher than source cluster, which might have compatibility issues
20211021:10:42:21 gpcopy:mxadmin:sdw7:026691-[ERROR]:-Caught an error: or: slice bounds out of range
20211021:10:42:21 gpcopy:mxadmin:sdw7:026691-[INFO]:-------------------------------------------------
20211021:10:42:21 gpcopy:mxadmin:sdw7:026691-[INFO]:-Total elapsed time: 953.799132ms
20211021:10:42:21 gpcopy:mxadmin:sdw7:026691-[INFO]:-Copied 0 databases

8、2 "gpcopy_temp_3" does not exist

failed to stop daemon on segments. pq: relation "gpcopy_temp_6" does not exist
Error: pq: relation "gpcopy_temp_3" does not exist

8、3 Control connection pair should call `InitConnections` first

20211025:13:32:01 gpcopy:mxadmin:sdw7:162638-[ERROR]:-Caught an error: or: slice bounds out of range [62:40]
20211025:13:32:01 gpcopy:mxadmin:sdw7:162638-[INFO]:-------------------------------------------------
20211025:13:32:01 gpcopy:mxadmin:sdw7:162638-[INFO]:-Total elapsed time: 74.028874ms
20211025:13:32:01 gpcopy:mxadmin:sdw7:162638-[INFO]:-Total transferred data 0B, transfer rate 0B/h
20211025:13:32:01 gpcopy:mxadmin:sdw7:162638-[INFO]:-Copied 0 databases
20211025:13:32:01 gpcopy:mxadmin:sdw7:162638-[WARNING]:-Encountered an error during cleanup: Control connection pair should call `InitConnections` first.

8、4 解决方式

1、检查greenplum和matrix的版本信息时候一致，因为gpcopy会校验版本

2、修改version信息需要在postgresql数据库下进行

3、替换一下红色的部门即可

CREATE OR REPLACE FUNCTION pg_catalog.version() RETURNS TEXT
AS $$
SELECT 'PostgreSQL 9.4.20 (Greenplum Database 6.0.0-beta.1 build commit:067fb8e43115bb087c872ee3d2269d869430263d) on x86_64-unknown-linux-gnu, compiled by gcc (GCC) 6.4.0, 64-bit compiled on Mar 19 2019 01:25:00';
$$ LANGUAGE SQL;

参考资料

MatrixDB - 时序数据库加载性能测试过程进入四维纵横yMatrix官网，体验超融合时序数据库MatrixDB。https://ymatrix.cn/blog/load-test-steps

MatrixDB - 时序数据库插入性能评测：MatrixDB是InfluxDB的78倍进入四维纵横yMatrix官网，体验超融合时序数据库MatrixDB。https://www.ymatrix.cn/blog/20210524-MatrixDB-insertperformance

以上是关于mxgate是gpcopy同步速度的2倍的主要内容，如果未能解决你的问题，请参考以下文章

SpringBoot+@Async注解一起用，速度提升100倍！

Tensorflow 2.0 Keras 的训练速度比 2.0 Estimator 慢 4 倍

倒带 30 秒音频并以 2 倍速度播放

视频倍速如何一键调整

HTML5 将 youtube 速度从 url 提高 2 倍？

结巴分词 java 高性能实现，是 huaban jieba 速度的 2倍