This is an automated email from the ASF dual-hosted git repository.
chengpan pushed a commit to branch branch-1.8
in repository https://gitbox.apache.org/repos/asf/kyuubi.git
The following commit(s) were added to refs/heads/branch-1.8 by this push:
new f86628f3c [KYUUBI #6018] Speed up GetTables operation for Spark
session catalog
f86628f3c is described below
commit f86628f3cc856265ff8214599836205d2cef2f5c
Author: Cheng Pan <[email protected]>
AuthorDate: Mon Jan 29 14:21:09 2024 +0800
[KYUUBI #6018] Speed up GetTables operation for Spark session catalog
# :mag: Description
## Issue References ๐
This pull request aims to speed up the GetTables operation for the Spark
session catalog.
As reported in https://github.com/apache/kyuubi/discussions/4956,
https://github.com/apache/kyuubi/discussions/5949, the GetTables operation is
quite slow in some cases, and in https://github.com/apache/kyuubi/pull/4444,
`kyuubi.operation.getTables.ignoreTableProperties` was introduced to speed up
the V2 catalog, but not covers session catalog.
## Describe Your Solution ๐ง
Extend the scope of `kyuubi.operation.getTables.ignoreTableProperties` to
cover the GetTables operation for the Spark session catalog.
Currently, the basic step of GetTables in the Spark engine is
```
val catalog: String = getCatalog(spark, catalogName)
val databases: Seq[String] = sessionCatalog.listDatabases(schemaPattern)
val identifiers: Seq[TableIdentifier] = catalog.listTables(db,
tablePattern, includeLocalTempViews = false)
val tableObjects: Seq[CatalogTable] = catalog.getTablesByName(identifiers)
```
then filter `tableObjects` with `tableTypes: Set[String]`.
The cost of `catalog.getTablesByName(identifiers)` is quite high when the
table number is large, e.g. dozen thousand.
For some cases, listing tables only for table name display, it is worth
speeding up the operation while ignoring some properties(e.g. table comments)
and query criteria(specifically in this case, when
`kyuubi.operation.getTables.ignoreTableProperties=true`, criteria `tableTypes`
will be ignored, and all tables and views will be treated as TABLE to return.)
## Types of changes :bookmark:
- [ ] Bugfix (non-breaking change which fixes an issue)
- [x] New feature (non-breaking change which adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
## Test Plan ๐งช
Pass GA
---
# Checklist ๐
- [x] This patch was not authored or co-authored using [Generative
Tooling](https://www.apache.org/legal/generative-tooling.html)
**Be nice. Be informative.**
Closes #6018 from pan3793/fast-get-table.
Closes #6018
058001c6f [Cheng Pan] fix
405b12484 [Cheng Pan] fix
615b7470f [Cheng Pan] Speed up GetTables operation
Authored-by: Cheng Pan <[email protected]>
Signed-off-by: Cheng Pan <[email protected]>
(cherry picked from commit d474768d97e85218183c880528a992a2dc258229)
Signed-off-by: Cheng Pan <[email protected]>
---
docs/configuration/settings.md | 2 +-
.../engine/spark/util/SparkCatalogUtils.scala | 37 ++++++++++++++++------
.../org/apache/kyuubi/config/KyuubiConf.scala | 3 +-
3 files changed, 30 insertions(+), 12 deletions(-)
diff --git a/docs/configuration/settings.md b/docs/configuration/settings.md
index 265d89795..69d220337 100644
--- a/docs/configuration/settings.md
+++ b/docs/configuration/settings.md
@@ -376,7 +376,7 @@ You can configure the Kyuubi properties in
`$KYUUBI_HOME/conf/kyuubi-defaults.co
| Key |
Default |
Meaning
[...]
|--------------------------------------------------|---------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
[...]
-| kyuubi.operation.getTables.ignoreTableProperties | false
| Speed up the
`GetTables` operation by returning table identities only.
[...]
+| kyuubi.operation.getTables.ignoreTableProperties | false
| Speed up the
`GetTables` operation by ignoring `tableTypes` query criteria, and returning
table identities only.
[...]
| kyuubi.operation.idle.timeout | PT3H
| Operation will be
closed when it's not accessed for this duration of time
[...]
| kyuubi.operation.interrupt.on.cancel | true
| When true, all running
tasks will be interrupted if one cancels a query. When false, all running tasks
will remain until finished.
[...]
| kyuubi.operation.language | SQL
| Choose a programing
language for the following inputs<ul><li>SQL: (Default) Run all following
statements as SQL queries.</li><li>SCALA: Run all following input as scala
codes</li><li>PYTHON: (Experimental) Run all following input as Python codes
with Spark engine</li></ul>
[...]
diff --git
a/externals/kyuubi-spark-sql-engine/src/main/scala/org/apache/kyuubi/engine/spark/util/SparkCatalogUtils.scala
b/externals/kyuubi-spark-sql-engine/src/main/scala/org/apache/kyuubi/engine/spark/util/SparkCatalogUtils.scala
index 18a14494e..b55319830 100644
---
a/externals/kyuubi-spark-sql-engine/src/main/scala/org/apache/kyuubi/engine/spark/util/SparkCatalogUtils.scala
+++
b/externals/kyuubi-spark-sql-engine/src/main/scala/org/apache/kyuubi/engine/spark/util/SparkCatalogUtils.scala
@@ -163,8 +163,8 @@ object SparkCatalogUtils extends Logging {
val namespaces = listNamespacesWithPattern(catalog, schemaPattern)
catalog match {
case builtin if builtin.name() == SESSION_CATALOG =>
- val catalog = spark.sessionState.catalog
- val databases = catalog.listDatabases(schemaPattern)
+ val sessionCatalog = spark.sessionState.catalog
+ val databases = sessionCatalog.listDatabases(schemaPattern)
def isMatchedTableType(tableTypes: Set[String], tableType: String):
Boolean = {
val typ = if (tableType.equalsIgnoreCase(VIEW)) VIEW else TABLE
@@ -172,22 +172,39 @@ object SparkCatalogUtils extends Logging {
}
databases.flatMap { db =>
- val identifiers = catalog.listTables(db, tablePattern,
includeLocalTempViews = false)
- catalog.getTablesByName(identifiers)
- .filter(t => isMatchedTableType(tableTypes, t.tableType.name)).map
{ t =>
- val typ = if (t.tableType.name == VIEW) VIEW else TABLE
+ val identifiers =
+ sessionCatalog.listTables(db, tablePattern, includeLocalTempViews
= false)
+ if (ignoreTableProperties) {
+ identifiers.map { ti: TableIdentifier =>
Row(
catalogName,
- t.database,
- t.identifier.table,
- typ,
- t.comment.getOrElse(""),
+ ti.database.getOrElse("default"),
+ ti.table,
+ TABLE, // ignore tableTypes criteria and simply treat all
table type as TABLE
+ "",
null,
null,
null,
null,
null)
}
+ } else {
+ sessionCatalog.getTablesByName(identifiers)
+ .filter(t => isMatchedTableType(tableTypes,
t.tableType.name)).map { t =>
+ val typ = if (t.tableType.name == VIEW) VIEW else TABLE
+ Row(
+ catalogName,
+ t.database,
+ t.identifier.table,
+ typ,
+ t.comment.getOrElse(""),
+ null,
+ null,
+ null,
+ null,
+ null)
+ }
+ }
}
case tc: TableCatalog =>
val tp = tablePattern.r.pattern
diff --git
a/kyuubi-common/src/main/scala/org/apache/kyuubi/config/KyuubiConf.scala
b/kyuubi-common/src/main/scala/org/apache/kyuubi/config/KyuubiConf.scala
index 70a7e65c4..b22a5131f 100644
--- a/kyuubi-common/src/main/scala/org/apache/kyuubi/config/KyuubiConf.scala
+++ b/kyuubi-common/src/main/scala/org/apache/kyuubi/config/KyuubiConf.scala
@@ -3226,7 +3226,8 @@ object KyuubiConf {
val OPERATION_GET_TABLES_IGNORE_TABLE_PROPERTIES: ConfigEntry[Boolean] =
buildConf("kyuubi.operation.getTables.ignoreTableProperties")
- .doc("Speed up the `GetTables` operation by returning table identities
only.")
+ .doc("Speed up the `GetTables` operation by ignoring `tableTypes` query
criteria, " +
+ "and returning table identities only.")
.version("1.8.0")
.booleanConf
.createWithDefault(false)
