Citus Research Reports

于 2022-03-30 00:57:23 发布
阅读量293 收藏
点赞数
分类专栏： postgres 文章标签： citus postgresql pgsql
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/u012731379/article/details/123835821
版权

Incomplete, More details are being added to the report

Hardware resources usage report

hash mode vs chunk mode report

devops on exists table vs non-exists table report

pgsql streaming replication vs citus replication Frequently Asked Questions - Citus 9.5 documentation

Prepare

The hardware configuration

server *3

cpu：4core memory：16G disk: hdd 300g

software configuration

Citus Plugin For Pgsql - Quickstart

test data

stream_616d2986bd17682e66aedbb3 899035

stream_617fc79a355b7f54684a9c8b 2693935

Sql
stream_616d2986bd17682e66aedbb3
default
explain analyse
SELECT
  "col0" AS diffcol0,
  count('*') AS count
FROM
  "stream_616d2986bd17682e66aedbb3" AS tab0,
  (
    SELECT
      tab0._id,
      UNNEST("f_7d1_ff35_t_labels$textual") AS "col0"
    FROM
      "stream_616d2986bd17682e66aedbb3" as tab0
  ) AS tab1
WHERE
  ("tab1"."_id" = "tab0"."_id")
GROUP BY
  "diffcol0"
ORDER BY
  "count" DESC NULLS LAST
LIMIT
  50 OFFSET 0;
ngram
explain analyse
SELECT
  "col0" AS diffcol0,
  count('*') AS count
FROM
  "stream_616d2986bd17682e66aedbb3" AS tab0,
  (
    SELECT
      tab0._id,
      c_unnest_taxonomy_with_null("f_7d1_ff35_t_labels$textual") AS "col0"
    FROM
      "stream_616d2986bd17682e66aedbb3" as tab0
  ) AS tab1
WHERE
  (
    (
      "f_n61_af35_ngrams$textual" && (ARRAY ['日期_新鲜','值得_信赖']::text [])
    )
    AND "tab1"."_id" = "tab0"."_id"
  )
GROUP BY
  "diffcol0"
ORDER BY
  "diffcol0" ASC NULLS FIRST
LIMIT
  200 OFFSET 0;
stream_617fc79a355b7f54684a9c8b
default
explain analyse
SELECT
  "col0" AS diffcol0,
  count('*') AS count
FROM
  "stream_617fc79a355b7f54684a9c8b" AS tab0,
  (
    SELECT
      tab0._id,
      UNNEST("f_jq2_20d7_t_labels$textual") AS "col0"
    FROM
      "stream_617fc79a355b7f54684a9c8b" as tab0
  ) AS tab1
WHERE
  ("tab1"."_id" = "tab0"."_id")
GROUP BY
  "diffcol0"
ORDER BY
  "count" DESC NULLS LAST
LIMIT
  50 OFFSET 0;
ngram
explain analyse
SELECT
  "col0" AS diffcol0,
  count('*') AS count
FROM
  "stream_617fc79a355b7f54684a9c8b" AS tab0,
  (
    SELECT
      tab0._id,
      c_unnest_taxonomy_with_null("f_jq2_20d7_t_labels$textual") AS "col0"
    FROM
      "stream_617fc79a355b7f54684a9c8b" as tab0
  ) AS tab1
WHERE
  (
    (
      "f_fq2_7de8_ngrams$textual" && (ARRAY ['日期_新鲜','值得_信赖']::text [])
    )
    AND "tab1"."_id" = "tab0"."_id"
  )
GROUP BY
  "diffcol0"
ORDER BY
  "diffcol0" ASC NULLS FIRST
LIMIT
  200 OFFSET 0;

testing

Hash mode

citus configure

set citus.shard_count=8;

You can set the number of global shards for the current database.

Default is 32. This parameter is valid only for hash mode

query speed results
Cluster mode report ( no backup)

set citus.shard_replication_factor=1;

Default is 1. No Fragment backup

0.9mclustershard4shard8shard16shard32
default3940 ms1833 ms1713 ms1558 ms (39.54%)1522 ms (38.62%)
ngram5406 ms1714 ms1152 ms893 ms (16.51%)974 ms (18.01%)
2.7mclustershard4shard8shard16shard32
default5719 ms2763 ms3406 ms2750 ms (48.08%)2881 ms (50.37%)
ngram9395 ms2757 ms2214 ms1843 ms (19.61%)1772 ms (18.86%)
Cluster mode report (backup)

set citus.shard_replication_factor=2;

Each shard has a backup

0.9mclustershard 4shard 8shard 16shard 32
default3940 ms2020 ms1910 ms1426 ms (36.19%)1327 ms (33.68%)
ngram5406 ms1617 ms863 ms920 ms(17.01%)996 ms(18.42%)
2.7mclustershard 4shard 8shard 16shard 32
default5719 ms2921 ms3126 ms2384 ms (41.68%)2533 ms (44.29%)
ngram9395 ms2769 ms2090 ms1710 ms (18.20%)1794 ms (19.09)
Cluster mode report(no backup + backup)
0.9mno backup (default)backup (default)no backup(ngram)backup (ngram)
shard 41833 ms2020 ms1671 ms1617 ms
shard 81713 ms1910 ms1152 ms863 ms
shard 161558 ms1426 ms893 ms920 ms
shard 321522 ms1327 ms974 ms996 ms
2.7mno backup (default)backup (default)no backup(ngram)backup (ngram)
shard 42763 ms2921 ms2757 ms2769 ms
shard 83406 ms3126 ms2214 ms2090 ms
shard 162750 ms2384 ms1843 ms1710 ms
shard 322881 ms2533 ms1772 ms1794 ms
Stand-alone mode report

Standalone does not require testing

conclusion

Based on the table above, we can see that 16 shards work best

cpu_core_number * (2-4) * number of nodes

shard backup test

The number of citus worker nodes is 3

The number of shard is 5

worker1（shard table id）worker2（shard table id）worker3 （shard table id）
chunk-102087chunk-102088chunk-102087
chunk-102088chunk-102089chunk-102089
chunk-102090chunk-102091chunk-102090
chunk-102091

The number of citus worker nodes is 3

The number of shard is 6

worker1（shard table id）worker2（shard table id）worker3 （shard table id）
chunk-102092chunk-102093chunk-102092
chunk-102093chunk-102094chunk-102094
chunk-102095chunk-102096chunk-102095
chunk-102096chunk-102097chunk-102097
Citus shardbackup failover test

citus worker nodes: 3

shard: 6

backup 2

QUERY PLAN

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 Limit  (cost=510.14..510.27 rows=50 width=40) (actual time=1043.209..1043.220 rows=50 loops=1)
   ->  Sort  (cost=510.14..510.64 rows=200 width=40) (actual time=1043.207..1043.210 rows=50 loops=1)
         Sort Key: (COALESCE((pg_catalog.sum(remote_scan.count))::bigint, '0'::bigint)) DESC NULLS LAST
         Sort Method: top-N heapsort  Memory: 30kB
         ->  HashAggregate  (cost=500.00..503.50 rows=200 width=40) (actual time=1043.111..1043.158 rows=163 loops=1)
               Group Key: remote_scan.diffcol0
               ->  Custom Scan (Citus Adaptive)  (cost=0.00..0.00 rows=100000 width=40) (actual time=1042.721..1042.791 rows=954 loops=1)
                     Task Count: 6
                     Tuple data received from nodes: 13 kB
                     Tasks Shown: One of 6
                     ->  Task
                           Tuple data received from node: 2171 bytes
                           Node: host=10.163.16.167 port=5434 dbname=db_61b061f8889c8eb9da8fe0d5
                           ->  Finalize GroupAggregate  (cost=101649.93..101700.60 rows=200 width=40) (actual time=1022.072..1022.481 rows=157 loops=1)
                           .......

 Planning Time: 3.273 ms
 Execution Time: 1043.295 ms

You can see from the above code that host=10.163.16.167. Manually shut down the Citus worker node.

QUERY PLAN

------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
----------------------
 Limit  (cost=510.14..510.27 rows=50 width=40) (actual time=1879.004..1879.014 rows=50 loops=1)
   ->  Sort  (cost=510.14..510.64 rows=200 width=40) (actual time=1879.001..1879.004 rows=50 loops=1)
         Sort Key: (COALESCE((pg_catalog.sum(remote_scan.count))::bigint, '0'::bigint)) DESC NULLS LAST
         Sort Method: top-N heapsort  Memory: 30kB
         ->  HashAggregate  (cost=500.00..503.50 rows=200 width=40) (actual time=1878.908..1878.955 rows=163 loops=1)
               Group Key: remote_scan.diffcol0
               ->  Custom Scan (Citus Adaptive)  (cost=0.00..0.00 rows=100000 width=40) (actual time=1878.567..1878.629 rows=954 loops=1)
                     Task Count: 6
                     Tuple data received from nodes: 13 kB
                     Tasks Shown: One of 6
                     ->  Task
                           Tuple data received from node: 2179 bytes
                           Node: host=10.163.21.120 port=5434 dbname=db_61b061f8889c8eb9da8fe0d5
                           ->  Finalize GroupAggregate  (cost=111112.90..111163.57 rows=200 width=40) (actual time=1803.301..1803.372 rows=158 loops=1)
                           ........

 Planning Time: 1.486 ms
 Execution Time: 1879.072 ms
(34 rows)

WARNING: connection to the remote node 10.163.16.167:5434 failed with the following error: server closed the connection unexpectedly
This probably means the server terminated abnormally
before or while processing the request.

After the worker node is closed manually, a warning will be displayed indicating that the query speed is not guaranteed, but data integrity is guaranteed.

citus worker node 4

shard 5

backup 2

Manually shut down two Citus worker nodes at random, the query is likely to fail, and the query is likely to succeed.

rebalance shards test
db_61b061f8889c8eb9da8fe0d5=# /*NO LOAD BLANCE*/ SELECT rebalance_table_shards('stream_616d2986bd17682e66aedbb3');
ERROR:  rebalance_table_shards() is only supported on Citus Enterprise

To use this feature, you need to upgrade to citus version 10.2, but citus10.2 supports postgresql-12 at a minimum.

Shard rebalancing is available in Citus Community edition, but shards are blocked for write access while being moved. For non-blocking reads and writes during rebalancing, try Citus Enterprise edition.Cluster Management - Citus 10.2 documentation

Disk usage test
0.9mtotalshard 4shard8
no backup2094MB524 + 524 + 522 + 524 （2094MB）263+262+262+261+262+263+262+262 (2097MB)
backup2094MB523+524+523+524+524+524+523+523 (4188MB)262+262+263+262+262+262+261+262+262+263+262+262+262+261+263+263 (4194MB)

The disk usage in PSQL primary and secondary modes is the same
These values are the size of each shard in MB

conclusion

According to the above table, several fragments can be created and evenly distributed to several worker nodes. Using backup will double the disk usage of worker nodes.

table_total_size / number of nodes * citus.shard_replication_factor

append mode

citus configure

select create_distributed_table('stream_616d2986bd17682e66aedbb3','_id', 'hash')

The Append mode requires an empty table to execute

set citus.shard_max_size="100MB";

Set the size of each shard and create a new shard.

create distributed tables
test_append=# select create_distributed_table('stream_616d2986bd17682e66aedbb3','_id', 'append'); 
WARNING:  table "stream_616d2986bd17682e66aedbb3" has a UNIQUE or EXCLUDE constraint 
DETAIL:  UNIQUE constraints, EXCLUDE constraints, and PRIMARY KEYs on append-partitioned tables cannot be enforced. 
HINT:  Consider using hash partitioning.  
create_distributed_table 
--------------------------
Execute SQL：
ERROR: cannot push down this subquery
DETAIL: Currently append partitioned relations with overlapping shard intervals are not supported

Description The query failed because primary_key was used as the distributed column

test again:
test_append=# select create_distributed_table('stream_616d2986bd17682e66aedbb3','f_vc_dbb3_Date$temporal', 'append');
WARNING:  table "stream_616d2986bd17682e66aedbb3" has a UNIQUE or EXCLUDE constraint
描述:  UNIQUE constraints, EXCLUDE constraints, and PRIMARY KEYs on append-partitioned tables cannot be enforced.
提示:  Consider using hash partitioning.
ERROR:  cannot create constraint on "stream_616d2986bd17682e66aedbb3"
描述:  Distributed relations cannot have UNIQUE, EXCLUDE, or PRIMARY KEY constraints that do not include the partition column (with an equality operator if EXCLUDE).

The append mode was not tested successfully
Append Distribution - Citus 9.5 documentation

Citus with PGSQL one master and two slave

Architecture diagram

在这里插入图片描述

The implementation process

  1. Install the Citus plug-in in the existing cluster，adding a configuration and restart

  2. Add citus worker node to master node.

    /*NO LOAD BLANCE*/ SELECT * from master_add_node('worker1_ip', 5432);

  3. Create a distributed table

    /*NO LOAD BLANCE*/ select create_distributed_table('stream_616d2986bd17682e66aedbb3','_id', 'hash');

  4. Validation

    SELECT * FROM master_get_active_worker_nodes();

SQL validation
Use citus to query
db_61b061f8889c8eb9da8fe0d5=# explain analyse
SELECT
  "col0" AS diffcol0,
  count('*') AS count
FROM
  "stream_616d2986bd17682e66aedbb3" AS tab0,
  (
    SELECT
      tab0._id,
      UNNEST("f_7d1_ff35_t_labels$textual") AS "col0"
    FROM
      "stream_616d2986bd17682e66aedbb3" as tab0
  ) AS tab1
WHERE
  ("tab1"."_id" = "tab0"."_id")
GROUP BY
  "diffcol0"
ORDER BY
  "count" DESC NULLS LAST
LIMIT
  50 OFFSET 0;
                                                                                                           QUERY PLAN
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 Limit  (cost=510.14..510.27 rows=50 width=40) (actual time=1038.773..1038.784 rows=50 loops=1)
   ->  Sort  (cost=510.14..510.64 rows=200 width=40) (actual time=1038.772..1038.774 rows=50 loops=1)
         Sort Key: (COALESCE((pg_catalog.sum(remote_scan.count))::bigint, '0'::bigint)) DESC NULLS LAST
         Sort Method: top-N heapsort  Memory: 30kB
         ->  HashAggregate  (cost=500.00..503.50 rows=200 width=40) (actual time=1038.682..1038.728 rows=163 loops=1)
               Group Key: remote_scan.diffcol0
               ->  Custom Scan (Citus Adaptive)  (cost=0.00..0.00 rows=100000 width=40) (actual time=1038.332..1038.415 rows=954 loops=1)
                     Task Count: 6
                     Tuple data received from nodes: 13 kB
                     Tasks Shown: One of 6
                     ->  Task
                           Tuple data received from node: 2179 bytes
                           Node: host=10.163.16.167 port=5434 dbname=db_61b061f8889c8eb9da8fe0d5
                           ->  Finalize GroupAggregate  (cost=102035.79..102086.46 rows=200 width=40) (actual time=1001.646..1001.941 rows=158 loops=1)
                                 Group Key: (unnest(tab0_1."f_7d1_ff35_t_labels$textual"))
                                 ->  Gather Merge  (cost=102035.79..102082.46 rows=400 width=40) (actual time=1001.600..1005.658 rows=467 loops=1)
                                       Workers Planned: 2
                                       Workers Launched: 2
                                       ->  Sort  (cost=101035.76..101036.26 rows=200 width=40) (actual time=993.660..993.673 rows=156 loops=3)
                                             Sort Key: (unnest(tab0_1."f_7d1_ff35_t_labels$textual"))
                                             Sort Method: quicksort  Memory: 36kB
                                             Worker 0:  Sort Method: quicksort  Memory: 36kB
                                             Worker 1:  Sort Method: quicksort  Memory: 36kB
                                             ->  Partial HashAggregate  (cost=101026.12..101028.12 rows=200 width=40) (actual time=992.773..992.805 rows=156 loops=3)
                                                   Group Key: (unnest(tab0_1."f_7d1_ff35_t_labels$textual"))
                                                   ->  Parallel Hash Join  (cost=38198.95..100793.91 rows=46442 width=32) (actual time=162.253..803.286 rows=372815 loops=3)
                                                         Hash Cond: (tab0_1._id = tab0._id)
                                                         ->  ProjectSet  (cost=0.00..61187.74 rows=4644200 width=36) (actual time=0.135..448.785 rows=372815 loops=3)
                                                               ->  Parallel Seq Scan on stream_616d2986bd17682e66aedbb3_102096 tab0_1  (cost=0.00..37618.43 rows=46442 width=36) (actual time=0.024..121.193 rows=49965 loops=3)
                                                         ->  Parallel Hash  (cost=37618.43..37618.43 rows=46442 width=4) (actual time=151.849..151.849 rows=49965 loops=3)
                                                               Buckets: 262144 (originally 131072)  Batches: 1 (originally 1)  Memory Usage: 8960kB
                                                               ->  Parallel Seq Scan on stream_616d2986bd17682e66aedbb3_102096 tab0  (cost=0.00..37618.43 rows=46442 width=4) (actual time=0.034..116.506 rows=49965 loops=3)
                               Planning Time: 1.457 ms
                               Execution Time: 1006.036 ms
 Planning Time: 1.402 ms
 Execution Time: 1038.842 ms
(36 rows)
Native queries
db_61b061f8889c8eb9da8fe0d5=# explain analyse
SELECT
  "col0" AS diffcol0,
  count('*') AS count
FROM
  "stream_616d2986bd17682e66aedbb3" AS tab0,
  (
    SELECT
      tab0._id,
      UNNEST("f_7d1_ff35_t_labels$textual") AS "col0"
    FROM
      "stream_616d2986bd17682e66aedbb3" as tab0
  ) AS tab1
WHERE
  ("tab1"."_id" = "tab0"."_id")
GROUP BY
  "diffcol0"
ORDER BY
  "count" DESC NULLS LAST
LIMIT
  50 OFFSET 0;
                                                                                                   QUERY PLAN

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
 Limit  (cost=671257.48..671257.60 rows=50 width=40) (actual time=3920.311..3991.489 rows=50 loops=1)
   ->  Sort  (cost=671257.48..671257.98 rows=200 width=40) (actual time=3920.310..3920.312 rows=50 loops=1)
         Sort Key: (count('*')) DESC NULLS LAST
         Sort Method: top-N heapsort  Memory: 31kB
         ->  Finalize GroupAggregate  (cost=671200.16..671250.83 rows=200 width=40) (actual time=3919.948..3920.268 rows=163 loops=1)
               Group Key: tab1.col0
               ->  Gather Merge  (cost=671200.16..671246.83 rows=400 width=40) (actual time=3919.939..3990.818 rows=480 loops=1)
                     Workers Planned: 2
                     Workers Launched: 2
                     ->  Sort  (cost=670200.14..670200.64 rows=200 width=40) (actual time=3914.206..3914.215 rows=160 loops=3)
                           Sort Key: tab1.col0
                           Sort Method: quicksort  Memory: 36kB
                           Worker 0:  Sort Method: quicksort  Memory: 36kB
                           Worker 1:  Sort Method: quicksort  Memory: 36kB
                           ->  Partial HashAggregate  (cost=670190.49..670192.49 rows=200 width=40) (actual time=3913.414..3913.436 rows=160 loops=3)
                                 Group Key: tab1.col0
                                 ->  Parallel Hash Join  (cost=228747.38..668317.52 rows=374595 width=32) (actual time=2950.412..3465.694 rows=2225782 loops=3)
                                       Hash Cond: (tab1._id = tab0._id)
                                       ->  Subquery Scan on tab1  (cost=0.00..421698.18 rows=899027 width=36) (actual time=0.051..1927.125 rows=2225782 loops=3)
                                             ->  ProjectSet  (cost=0.00..412707.91 rows=37459500 width=36) (actualtime=0.048..1614.679 rows=2225782 loops=3)
                                                   ->  Parallel Seq Scan on stream_616d2986bd17682e66aedbb3 tab0_1 (cost=0.00..222600.95 rows=374595 width=91) (actual time=0.033..476.560 rows=299678 loops=3)
                                       ->  Parallel Hash  (cost=222600.95..222600.95 rows=374595 width=4) (actual time=527.937..527.938 rows=299678 loops=3)
                                             Buckets: 131072  Batches: 16  Memory Usage: 3264kB
                                             ->  Parallel Seq Scan on stream_616d2986bd17682e66aedbb3 tab0  (cost=0.00..222600.95 rows=374595 width=4) (actual time=0.040..455.306 rows=299678 loops=3)
 Planning Time: 0.390 ms
 Execution Time: 3991.600 ms
(26 rows)
zhipeng-python
关注
  • 0
    点赞
  • 踩
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
Citus 分布式 PostgreSQL 集群 - SQL Reference(SQL支持和变通方案)
o__cc的博客
04-02 430
由于 Citus 通过扩展 PostgreSQL 提供分布式功能，因此它与 PostgreSQL 结构兼容。这意味着用户可以使用丰富且可扩展的 PostgreSQL 生态系统附带的工具和功能来处理使用 Citus 创建的分布式表。 Citus 对它能够在单个工作节点上执行的任何查询具有 100% 的 SQL 覆盖率。 在访问有关单个租户的信息时，此类查询在多租户应用程序中很常见。 甚至跨节点查...
PostgreSQL - citus如何实现高可用，处理单点故障
研究与专注
01-15 3409
目录 1. worker节点故障 1.1 PostgreSQL的流式复制 1.2 Citus的分片复制 2. Coordinator节点故障 3. 总结 号外：Citus发布了8.x版本，支持PostgreSQL11。 Citus适合存放数据量较大的情形，不上亿的数据往Citus中存意义并不大，更适合放单节点。而存储如此大量的数据，如果不做好数据备份或者高可用，数据丢失产生的后果...
postgresql在使用citus添加其他服务器节点时报错：ERROR: connection to the remote node failed with the following...
fanfanfan的博客
06-09 1561
错误：无法访问远程服务器 解决： 1、主服务器和从服务器都需要配置允许远程访问：pg_hbf.conf文件中# IPv4 local connections字段，0.0.0.0/0代表允许所有服务器访问 2、主服务器和从服务器都需要配置监听以及端口：postgresql.conf文件中listen_addresses = '*'（代表监听所有访问的ip）字段以及port=5432字段 3、开启5432端口的防火墙 如果5432不在防火墙列表中，那么外部也是访问不到该服务器的5432端口..
Citus中的分片策略：Append Distribution追加分配
皓月如我的专栏
03-19 2652
翻译自《Citus官方文档》 Append Distribution追加分配 Append distribution is a specialized technique which requires care to use efficiently. Hash distribution is a better choice for most situations. 追加分配是需要谨慎使用...
Citus分布式方案（七）- 集群管理
印府的菜园子
09-09 3241
目录（七）集群管理选择集群的大小1. 分片的数量2. 多租户SaaS3. 实时分析初始硬件大小1. 多租户SaaS2. 实时分析集群扩容1. 添加一个数据节点2. 在不停机的情况下重新平衡分片3. 工作原理4. 增加一个协调节点（开发中） （七）集群管理 在本节内容：从Citus集群中添加或删除节点，以及如何处理节点故障。 选择集群的大小 为了更容易跨节点移动分片或在失败节点上重新复制分片，Citus企业版提供了一个分片重均衡的扩展。在此之前，我们将了解下在生产环境中运行集群的配置。 1. 分片的数量 集群
citus_docs：Citus文档
02-04
适用于Citus文档的Sphinx源文件 生成HTML版本： 从安装Sphinx 克隆此存储库 产生HTML cd citus_docs make # open _build/index.html in your browser Sphinx安装注意：在OS X上，最好通过而非Homebrew安装sphinx...
citus-5.2.2.tar.gz
01-13
标题中的"citus-5.2.2.tar.gz"是一个压缩包文件，通常包含源代码或者软件的安装包。"citus"是这个包的核心部分，它是一个针对PostgreSQL（简称PG）数据库系统的分布式扩展。"5.2.2"则是这个版本的编号，表示这是...
citus扩容实践
12-14
分布式数据库中的扩容是必不可少的需求，citus企业版中可以进行动态扩容，社区版中尚不支持动态扩容。不过我们可以通过修改元数据表中的分片信息来实现类似的功能。 本次实验扩容前只有一个work节点，扩容成两个work...
docker:Citus的Docker映像和配置
05-15
Citus是基于PostgreSQL分布式RDBMS。 有关更多信息，请参见。 功能 该映像使用标准配置值提供了一个正在运行的Citus实例（在PostgreSQL 12.2之上）。 它基于，因此请务必参考该映像的文档以获取高级配置选项（包括...
citus：分布式PostgreSQL作为扩展-用于多租户和实时分析工作负载
01-30
什么是Citus？ 开源PostgreSQL扩展（不是fork） 专为跨多个节点而扩展 用于查询并行化的分布式引擎 数据库设计，向外扩展的多租户应用程序，实时分析仪表盘和高通量的事务性工作负载 Citus是Postgres的开源扩展，...
MySQL-优化-ICP(Index condition pushdown)详解
sunashe的个人随笔
10-29 1853
文章目录MySQL-优化-ICP(Index condition pushdown)详解一、关于ICP二、什么情况下可以用到MySQL ICP特性三、疑问：如下这个sql为什么会使用到ICP呢？四、代码控制server层处理engine层处理 MySQL-优化-ICP(Index condition pushdown)详解 一、关于ICP ICP是index condition pushdown的...
PostgreSQL 源码解读（54）- 查询语句#39（make_one_rel函数#4-生...
09-25 267
本节继续介绍make_one_rel函数中的set_base_rel_sizes及其子函数。在set_base_rel_sizes函数调用过程中，如RTE为子查询，则生成子查询的访问路径，通过调用函数s...
PostgreSQL 源码解读（28）- 查询语句#13（查询优化-上拉子链接#3）
08-27 692
本节简单介绍了PG查询逻辑优化中的子查询链接(subLink),以EXISTS子链接为例介绍了子查询链接上拉主函数处理逻辑以及使用gdb跟踪分析。 一、源码解读 上一节介绍了ANY子链接,本节介绍了EX...
OCP007 考题解析（51-80）
weixin_33924770的博客
10-16 628
Technorati 标签: oracle,ocp,007,题库 Q: 51 Examine the structure of the EMPLOYEES and NEW_EMPLOYEES tables: EMPLOYEES EMPLOYEE_ID NUMBER Primary Key FIRST_NAME...
sql 违反了 PRIMARY KEY 约束,不能在对象 中插入重复键
热门推荐
睦生
04-22 7万+
sql 违反了 PRIMARY KEY 约束,不能在对象 中插入重复键说明你的数据里面有重复记录两种情况1.已存在的表中和要导入数据之间的重复这个时候可以通过在两个表之间建立关联，将主键级联找出重复记录2.要导入的表中存在重复记录可通过类似如下的语句将表中的重复记录找出来select StudentCode from StudentScore group by StudentCode[此处为 PR...
Unique约束,Primary Key约束与索引的关系学习与测试_20091213.doc
12-13 155
Unique约束,Primary Key约束与索引的关系 <> P517 <> P528 1 Creating an Index Associated with a Constraint O...
Oracle 唯一 约束（unique constraint） 与 索引（index） 关系说明
qjoycn
07-13 710
一. 官网对Unique Constraints说明 http://download.oracle.com/docs/cd/E11882_01/server.112/e16508/datainte.htm#CNCPT1642 uniquekey constraint requires that every value in a column or set of columns beuniqu...
Citus分布式数据库入门与实战
"Citus入门与生产实践.pptx - 陈华军 - 苏宁云软件公司架构专家 - PostgreSQL分布式数据库介绍" 本文将深入探讨Citus，一个基于PostgreSQL的分布式数据库系统，以及它在实际生产环境中的应用。Citus解决传统单机...
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值