[fix](Nereids) fix some sql cache consistence bug between multiple frontends #33722
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 38734 ms |
TPC-DS: Total hot run time: 183943 ms |
ClickBench: Total hot run time: 30.64 s |
|
Load test result on machine: 'aliyun_ecs.c7a.8xlarge_32C64G' |
Contributor
Author
|
run buildall |
924060929
force-pushed
the
sql_cache_consistence
branch
from
46b8fec to
002dc70
Compare
Contributor
Author
|
run buildall |
1 similar comment
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 38283 ms |
TPC-DS: Total hot run time: 185121 ms |
ClickBench: Total hot run time: 30.01 s |
|
Load test result on machine: 'aliyun_ecs.c7a.8xlarge_32C64G' |
morrySnow
approved these changes
Apr 17, 2024
Contributor
|
PR approved by at least one committer and no changes requested. |
github-actions
bot
added
approved
Contributor
|
PR approved by anyone and no changes requested. |
yiguolei
requested changes
Apr 17, 2024
Contributor
yiguolei
left a comment
yiguolei
left a comment
There was a problem hiding this comment.
Sync between multi FEs is very difficult.
We could simply using consistency hash like memcache to solve the problem.
github-actions
bot
removed
the
approved
Contributor
Author
there is no any sync logic in this pr, and the consistency hash is not changed |
yiguolei
approved these changes
Apr 17, 2024
github-actions
bot
added
the
approved
Contributor
|
PR approved by at least one committer and no changes requested. |
924060929
added a commit
to 924060929/incubator-doris
that referenced
this pull request
Apr 18, 2024
…ontends (apache#33722) fix some sql cache consistence bug between multiple frontends which introduced by [enhancement](Nereids) Enable parse sql from sql cache apache#33262, fix by use row policy as the part of sql cache key. support dynamic update the num of fe manage sql cache key (cherry picked from commit 90abd76)
yiguolei
pushed a commit
that referenced
this pull request
Apr 18, 2024
…ugs (#33852) * [enhancement](Nereids) Enable parse sql from sql cache (#33262) Before this pr, the query must pass through parser, analyzer, rewriter, optimizer and translator, then we can check whether this query can use sql cache, if the query is too long, or the number of join tables too big, the plan time usually >= 500ms. This pr reduce this time by skip the fashion plan path, because we can reuse the previous physical plan and query result if no any changed. In some cases we should not parse sql from sql cache, e.g. table structure changed, data changed, user policies changed, privileges changed, contains non-deterministic functions, and user variables changed. In my test case: query a view which has lots of join and union, and the tables has empty partition, the query latency is about 3ms. if not parse sql from sql cache, the plan time is about 550ms ## Features 1. use Config.sql_cache_manage_num to control how many sql cache be reused in on fe 2. if explain plan appear some plans contains `LogicalSqlCache` or `PhysicalSqlCache`, it means the query can use sql cache, like this: ```sql mysql> set enable_sql_cache=true; Query OK, 0 rows affected (0.00 sec) mysql> explain physical plan select * from test.t; +----------------------------------------------------------------------------------+ | Explain String(Nereids Planner) | +----------------------------------------------------------------------------------+ | cost = 3.135 | | PhysicalResultSink[53] ( outputExprs=[c1#0, c2#1] ) | | +--PhysicalDistribute[50]@0 ( stats=3, distributionSpec=DistributionSpecGather ) | | +--PhysicalOlapScan[t]@0 ( stats=3 ) | +----------------------------------------------------------------------------------+ 4 rows in set (0.02 sec) mysql> select * from test.t; +------+------+ | c1 | c2 | +------+------+ | 1 | 2 | | -2 | -2 | | NULL | 30 | +------+------+ 3 rows in set (0.05 sec) mysql> explain physical plan select * from test.t; +-------------------------------------------------------------------------------------------+ | Explain String(Nereids Planner) | +-------------------------------------------------------------------------------------------+ | cost = 0.0 | | PhysicalSqlCache[2] ( queryId=78511f515cda466b-95385d892d6c68d0, backend=127.0.0.1:9050 ) | | +--PhysicalResultSink[52] ( outputExprs=[c1#0, c2#1] ) | | +--PhysicalDistribute[49]@0 ( stats=3, distributionSpec=DistributionSpecGather ) | | +--PhysicalOlapScan[t]@0 ( stats=3 ) | +-------------------------------------------------------------------------------------------+ 5 rows in set (0.01 sec) ``` (cherry picked from commit 03bd2a3) * fix * [fix](Nereids) fix some sql cache consistence bug between multiple frontends (#33722) fix some sql cache consistence bug between multiple frontends which introduced by [enhancement](Nereids) Enable parse sql from sql cache #33262, fix by use row policy as the part of sql cache key. support dynamic update the num of fe manage sql cache key (cherry picked from commit 90abd76) * [fix](Nereids) fix bug of dry run query with sql cache (#33799) 1. dry run query should not use sql cache 2. fix test sql cache in cloud mode 3. enable cache OneRowRelation and EmptyRelation in frontend to skip parse sql (cherry picked from commit dc80ecf) * remove cloud mode
924060929
added a commit
to 924060929/incubator-doris
that referenced
this pull request
Apr 19, 2024
…ontends (apache#33722) fix some sql cache consistence bug between multiple frontends which introduced by [enhancement](Nereids) Enable parse sql from sql cache apache#33262, fix by use row policy as the part of sql cache key. support dynamic update the num of fe manage sql cache key (cherry picked from commit 90abd76)
yiguolei
pushed a commit
that referenced
this pull request
Apr 19, 2024
…ugs (#33867) * [enhancement](Nereids) Enable parse sql from sql cache (#33262) Before this pr, the query must pass through parser, analyzer, rewriter, optimizer and translator, then we can check whether this query can use sql cache, if the query is too long, or the number of join tables too big, the plan time usually >= 500ms. This pr reduce this time by skip the fashion plan path, because we can reuse the previous physical plan and query result if no any changed. In some cases we should not parse sql from sql cache, e.g. table structure changed, data changed, user policies changed, privileges changed, contains non-deterministic functions, and user variables changed. In my test case: query a view which has lots of join and union, and the tables has empty partition, the query latency is about 3ms. if not parse sql from sql cache, the plan time is about 550ms ## Features 1. use Config.sql_cache_manage_num to control how many sql cache be reused in on fe 2. if explain plan appear some plans contains `LogicalSqlCache` or `PhysicalSqlCache`, it means the query can use sql cache, like this: ```sql mysql> set enable_sql_cache=true; Query OK, 0 rows affected (0.00 sec) mysql> explain physical plan select * from test.t; +----------------------------------------------------------------------------------+ | Explain String(Nereids Planner) | +----------------------------------------------------------------------------------+ | cost = 3.135 | | PhysicalResultSink[53] ( outputExprs=[c1#0, c2#1] ) | | +--PhysicalDistribute[50]@0 ( stats=3, distributionSpec=DistributionSpecGather ) | | +--PhysicalOlapScan[t]@0 ( stats=3 ) | +----------------------------------------------------------------------------------+ 4 rows in set (0.02 sec) mysql> select * from test.t; +------+------+ | c1 | c2 | +------+------+ | 1 | 2 | | -2 | -2 | | NULL | 30 | +------+------+ 3 rows in set (0.05 sec) mysql> explain physical plan select * from test.t; +-------------------------------------------------------------------------------------------+ | Explain String(Nereids Planner) | +-------------------------------------------------------------------------------------------+ | cost = 0.0 | | PhysicalSqlCache[2] ( queryId=78511f515cda466b-95385d892d6c68d0, backend=127.0.0.1:9050 ) | | +--PhysicalResultSink[52] ( outputExprs=[c1#0, c2#1] ) | | +--PhysicalDistribute[49]@0 ( stats=3, distributionSpec=DistributionSpecGather ) | | +--PhysicalOlapScan[t]@0 ( stats=3 ) | +-------------------------------------------------------------------------------------------+ 5 rows in set (0.01 sec) ``` (cherry picked from commit 03bd2a3) * fix * [fix](Nereids) fix some sql cache consistence bug between multiple frontends (#33722) fix some sql cache consistence bug between multiple frontends which introduced by [enhancement](Nereids) Enable parse sql from sql cache #33262, fix by use row policy as the part of sql cache key. support dynamic update the num of fe manage sql cache key (cherry picked from commit 90abd76) * [fix](Nereids) fix bug of dry run query with sql cache (#33799) 1. dry run query should not use sql cache 2. fix test sql cache in cloud mode 3. enable cache OneRowRelation and EmptyRelation in frontend to skip parse sql (cherry picked from commit dc80ecf) * remove cloud mode * remove @NotNull
yiguolei
pushed a commit
that referenced
this pull request
Apr 19, 2024
…ugs (#33867) * [enhancement](Nereids) Enable parse sql from sql cache (#33262) Before this pr, the query must pass through parser, analyzer, rewriter, optimizer and translator, then we can check whether this query can use sql cache, if the query is too long, or the number of join tables too big, the plan time usually >= 500ms. This pr reduce this time by skip the fashion plan path, because we can reuse the previous physical plan and query result if no any changed. In some cases we should not parse sql from sql cache, e.g. table structure changed, data changed, user policies changed, privileges changed, contains non-deterministic functions, and user variables changed. In my test case: query a view which has lots of join and union, and the tables has empty partition, the query latency is about 3ms. if not parse sql from sql cache, the plan time is about 550ms ## Features 1. use Config.sql_cache_manage_num to control how many sql cache be reused in on fe 2. if explain plan appear some plans contains `LogicalSqlCache` or `PhysicalSqlCache`, it means the query can use sql cache, like this: ```sql mysql> set enable_sql_cache=true; Query OK, 0 rows affected (0.00 sec) mysql> explain physical plan select * from test.t; +----------------------------------------------------------------------------------+ | Explain String(Nereids Planner) | +----------------------------------------------------------------------------------+ | cost = 3.135 | | PhysicalResultSink[53] ( outputExprs=[c1#0, c2#1] ) | | +--PhysicalDistribute[50]@0 ( stats=3, distributionSpec=DistributionSpecGather ) | | +--PhysicalOlapScan[t]@0 ( stats=3 ) | +----------------------------------------------------------------------------------+ 4 rows in set (0.02 sec) mysql> select * from test.t; +------+------+ | c1 | c2 | +------+------+ | 1 | 2 | | -2 | -2 | | NULL | 30 | +------+------+ 3 rows in set (0.05 sec) mysql> explain physical plan select * from test.t; +-------------------------------------------------------------------------------------------+ | Explain String(Nereids Planner) | +-------------------------------------------------------------------------------------------+ | cost = 0.0 | | PhysicalSqlCache[2] ( queryId=78511f515cda466b-95385d892d6c68d0, backend=127.0.0.1:9050 ) | | +--PhysicalResultSink[52] ( outputExprs=[c1#0, c2#1] ) | | +--PhysicalDistribute[49]@0 ( stats=3, distributionSpec=DistributionSpecGather ) | | +--PhysicalOlapScan[t]@0 ( stats=3 ) | +-------------------------------------------------------------------------------------------+ 5 rows in set (0.01 sec) ``` (cherry picked from commit 03bd2a3) * fix * [fix](Nereids) fix some sql cache consistence bug between multiple frontends (#33722) fix some sql cache consistence bug between multiple frontends which introduced by [enhancement](Nereids) Enable parse sql from sql cache #33262, fix by use row policy as the part of sql cache key. support dynamic update the num of fe manage sql cache key (cherry picked from commit 90abd76) * [fix](Nereids) fix bug of dry run query with sql cache (#33799) 1. dry run query should not use sql cache 2. fix test sql cache in cloud mode 3. enable cache OneRowRelation and EmptyRelation in frontend to skip parse sql (cherry picked from commit dc80ecf) * remove cloud mode * remove @NotNull
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Proposed changes