UPDATE · k8s原理與實踐

[TOC] # UPDATE UPDATE 算子用于更新數據表中滿足指定條件的數據行。 OceanBase 數據庫支持的 UPDATE 算子包括 UPDATE 和 MULTI PARTITION UPDATE。 ## UPDATE UPDATE 算子用于更新數據表單個分區中的數據。如下例所示，Q1 查詢更新了表 t1 中所有滿足`c2 = '100'`的行，并將 c2 的值設置為 200。 ~~~ obclient>CREATE TABLE t1 (c1 INT PRIMARY KEY, c2 VARCHAR2(10)); Query OK, 0 rows affected (0.12 sec) obclient>CREATE TABLE t2 (c1 INT PRIMARY KEY, c2 VARCHAR2(10)) PARTITION BY HASH(c1) PARTITIONS 10; Query OK, 0 rows affected (0.12 sec) obclient>CREATE TABLE t3 (c1 INT PRIMARY KEY, c2 VARCHAR2(10)); Query OK, 0 rows affected (0.12 sec) obclient>CREATE INDEX IDX_t3_c2 ON t3 (c2) PARTITION BY HASH(c2) PARTITIONS 3; Query OK, 0 rows affected (0.12 sec) Q1: obclient>EXPLAIN UPDATE t1 SET c2 = '200' WHERE c2 = '100'\G; *************************** 1. row *************************** Query Plan: ====================================== |ID|OPERATOR |NAME|EST. ROWS|COST | ----------------------------------------------- |0 |EXCHANGE IN REMOTE | |990 |109687| |1 | EXCHANGE OUT REMOTE| |990 |109687| |2 | UPDATE | |990 |109687| |3 | TABLE SCAN |T1 |990 |108697| =============================================== Outputs & filters: ------------------------------------- 0 - output(nil), filter(nil) 1 - output(nil), filter(nil) 2 - output(nil), filter(nil), table_columns([{T1: ({T1: (T1.C1, T1.C2)})}]), update([T1.C2=?]) 3 - output([T1.C1], [T1.C2], [?]), filter([T1.C2 = '100']), access([T1.C2], [T1.C1]), partitions(p0) ~~~ 上述示例中，執行計劃展示中的 outputs & filters 詳細列出了 UPDATE 算子的輸出信息如下： <table data-tag="table" id="table-4g2-dvt-h1n" class="table"><colgroup span="1" width="139" data-tag="col" id="col-6z3-a39-jkd" colwidth="1*" colnum="1" colname="col1" style="width:50%" class="col"></colgroup><colgroup span="1" width="609" data-tag="col" id="col-pbd-hag-xe1" colwidth="1*" colnum="2" colname="col2" style="width:50%" class="col"></colgroup><thead id="thead-5k7-wml-vb6" class="thead"><tr id="tr-slg-h02-z4q"><th id="td-wut-mgt-rb2">信息名稱</th><th id="td-44m-nc5-kun">含義</th></tr></thead><tbody data-tag="tbody" id="tbody-yiy-05s-qxt" class="tbody"><tr data-tag="tr" id="tr-wml-902-twm" class="tr"><td data-tag="td" id="td-bj7-8a2-ion" class="td">output</td><td data-tag="td" id="td-noa-4hj-4sf" class="td">該算子輸出的表達式。</td></tr><tr data-tag="tr" id="tr-7mr-bdc-o1t" class="tr"><td data-tag="td" id="td-m93-cjt-brb" class="td">filter</td><td data-tag="td" id="td-ocu-56t-oyv" class="td">該算子上的過濾條件。由于示例中 UPDATE 算子沒有 filter，所以為 nil。對于更新語句，WHERE 中的謂詞會下推到基表上，比如 Q1 查詢中的 <code data-tag="code" class="code">c2 = '100'</code> 被下推到了 1 號算子上。</td></tr><tr data-tag="tr" id="tr-6rx-8fr-qgj" class="tr"><td data-tag="td" id="td-irp-w6u-u2f" class="td">table_columns</td><td data-tag="td" id="td-2kq-bfw-3u0" class="td">更新操作涉及的數據表的列。</td></tr><tr data-tag="tr" id="tr-ene-s67-joy" class="tr"><td data-tag="td" id="td-6h7-ylb-6or" class="td">update</td><td data-tag="td" id="td-1sy-4wv-xfq" class="td">更新操作中所有的賦值表達式。</td></tr></tbody></table> 更多 UPDATE 算子的示例如下： * Q2 查詢更新 t1 中的所有數據行，并將 c2 的值置為 200。 * Q3 查詢更新分區表 t2 中滿足`c1='100'`的數據行，并將 c2 的值置為 150。 * Q4 查詢更新分區表 t2 中滿足`c2 ='100'`的數據行，并將`c2`的值置為`rpad(t2.c2, 10, '9')`。從執行計劃中可以看到，UPDATE 算子分配在 EXCHANGE 算子下面，因此 2 號和 3 號算子會作為一個 task 以分區的粒度進行調度。執行時 3 號算子掃描出 t2 一個分區中滿足`c2 = '100'`的數據，2 號 UPDATE 算子則只會更新相應分區下掃描出的數據。 ~~~ Q2: obclient>EXPLAIN UPDATE t1 SET c2 = '200'\G; *************************** 1. row *************************** Query Plan: ====================================== |ID|OPERATOR |NAME|EST. ROWS|COST | ----------------------------------------------- |0 |EXCHANGE IN REMOTE | |100000 |161860| |1 | EXCHANGE OUT REMOTE| |100000 |161860| |2 | UPDATE | |100000 |161860| |3 | TABLE SCAN |T1 |100000 |61860 | =============================================== Outputs & filters: ------------------------------------- 0 - output(nil), filter(nil) 1 - output(nil), filter(nil) 2 - output(nil), filter(nil), table_columns([{T1: ({T1: (T1.C1, T1.C2)})}]), update([T1.C2=?]) 3 - output([T1.C1], [T1.C2], [?]), filter(nil), access([T1.C2], [T1.C1]), partitions(p0) Q3: obclient>EXPLAIN UPDATE t2 SET t2.c2 = '150' WHERE t2.c1 = '100'\G; *************************** 1. row *************************** Query Plan: =================================== |ID|OPERATOR |NAME|EST. ROWS|COST| ----------------------------------- |0 |UPDATE | |1 |53 | |1 | TABLE GET|T2 |1 |52 | =================================== Outputs & filters: ------------------------------------- 0 - output(nil), filter(nil), table_columns([{T2: ({T2: (T2.C1, T2.C2)})}]), update([T2.C2=?]) 1 - output([T2.C1], [T2.C2], [?]), filter(nil), access([T2.C1], [T2.C2]), partitions(p5) Q4: obclient>EXPLAIN UPDATE t2 SET t2.c2 = RPAD(t2.c2, 10, '9') WHERE t2.c2 = '100'\G; *************************** 1. row *************************** Query Plan: =============================================== |ID|OPERATOR |NAME |EST. ROWS|COST | ------------------------------------------------------- |0 |PX COORDINATOR | |9900 |1096793| |1 | EXCHANGE OUT DISTR |:EX10000|9900 |1096793| |2 | PX PARTITION ITERATOR| |9900 |1096793| |3 | UPDATE | |9900 |1096793| |4 | TABLE SCAN |T2 |9900 |1086893| ======================================================= Outputs & filters: ------------------------------------- 0 - output(nil), filter(nil) 1 - output(nil), filter(nil), dop=1 2 - output(nil), filter(nil) 3 - output(nil), filter(nil), table_columns([{T2: ({T2: (T2.C1, T2.C2)})}]), update([T2.C2=column_conv(VARCHAR,utf8mb4_bin,length:10,NULL,RPAD(T2.C2, 10, ?))]) 4 - output([T2.C1], [T2.C2], [column_conv(VARCHAR,utf8mb4_bin,length:10,NULL,RPAD(T2.C2, 10, ?))]), filter([T2.C2 = '100']), access([T2.C1], [T2.C2]), partitions(p[0-9]) ~~~ ## MULTI PARTITION UPDATE MULTI PARTITION UPDATE 算子表示更新數據表多個分區中的數據。如下例所示，Q5 查詢更新表 t3 中所有滿足`c2 < '100'`的數據行，并將 c2 的值置為 200。雖然 t3 本身是一個非分區表，但 t3 上存在全局索引 idx\_t3\_c2，因此每一條數據行會存在于多個分區中。 ~~~ Q5: obclient>EXPLAIN UPDATE t3 SET c2 = '200' WHERE c2 < '100'\G; *************************** 1. row *************************** Query Plan: ======================================================== |ID|OPERATOR |NAME |EST. ROWS|COST | ----------------------------------------------------------- |0 |MULTI PARTITION UPDATE | |10001 |27780| |1 | PX COORDINATOR | |10001 |17780| |2 | EXCHANGE OUT DISTR |:EX10000 |10001 |14941| |3 | PX PARTITION ITERATOR| |10001 |14941| |4 | TABLE SCAN |T3(IDX_T3_C2)|10001 |14941| =========================================================== Outputs & filters: ------------------------------------- 0 - output(nil), filter(nil), table_columns([{T3: ({T3: (T3.C1, T3.C2)}, {IDX_T3_C2: (T3.C2, T3.C1)})}]), update([T3.C2=?]) 1 - output([T3.C1], [T3.C2], [?]), filter(nil) 2 - output([T3.C2], [T3.C1], [?]), filter(nil), dop=1 3 - output([T3.C2], [T3.C1], [?]), filter(nil) 4 - output([T3.C2], [T3.C1], [?]), filter(nil), access([T3.C2], [T3.C1]), partitions(p[0-2]) ~~~ 更多 MULTI PARTITION UPDATE 的示例如下： * Q6 查詢更新分區表 t2 中滿足`c1 = 100`的數據行，并將 c1 的值設置為 101。因為更新的列是主鍵列，可能會導致更新后的數據行與更新前的數據行位于不同的分區，因此需要使用 MULTI PARTITION UPDATE 算子進行更新。 ~~~ Q6: obclient>EXPLAIN UPDATE t2 SET t2.c1 = 101 WHERE t2.c1 = 100\G; *************************** 1. row *************************** Query Plan: =============================================== |ID|OPERATOR |NAME|EST. ROWS|COST| ----------------------------------------------- |0 |MULTI PARTITION UPDATE| |1 |54 | |1 | EXCHANGE IN DISTR | |1 |53 | |2 | EXCHANGE OUT DISTR | |1 |52 | |3 | TABLE GET |T2 |1 |52 | =============================================== Outputs & filters: ------------------------------------- 0 - output(nil), filter(nil), table_columns([{T2: ({T2: (T2.C1, T2.C2)})}]), update([T2.C1=?]) 1 - output([T2.C1], [T2.C2], [?]), filter(nil) 2 - output([T2.C1], [T2.C2], [?]), filter(nil) 3 - output([T2.C1], [T2.C2], [?]), filter(nil), access([T2.C1], [T2.C2]), partitions(p5) ~~~