[
https://issues.apache.org/jira/browse/PHOENIX-6767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17585187#comment-17585187
]
ASF GitHub Bot commented on PHOENIX-6767:
-----------------------------------------
stoty commented on code in PR #1493:
URL: https://github.com/apache/phoenix/pull/1493#discussion_r955727536
##########
phoenix-core/src/main/java/org/apache/phoenix/iterate/BaseResultIterators.java:
##########
@@ -929,12 +929,19 @@ private static boolean clipKeyRangeBytes(RowKeySchema
schema, int fieldIndex, in
* @throws SQLException
*/
private List<List<Scan>> getParallelScans(byte[] startKey, byte[] stopKey)
throws SQLException {
- List<HRegionLocation> regionLocations =
getRegionBoundaries(scanGrouper);
- List<byte[]> regionBoundaries = toBoundaries(regionLocations);
ScanRanges scanRanges = context.getScanRanges();
PTable table = getTable();
- boolean isSalted = table.getBucketNum() != null;
boolean isLocalIndex = table.getIndexType() == IndexType.LOCAL;
+ if(!isLocalIndex && scanRanges.isPointLookup()) {
Review Comment:
nit:This is rightfully flagged by checkstyle
##########
phoenix-core/src/it/java/org/apache/phoenix/end2end/salted/SaltedTableIT.java:
##########
@@ -62,6 +62,33 @@ public void testTableWithInvalidBucketNumber() throws
Exception {
}
}
+ @Test
+ public void testPointLookupOnSaltedTable() throws Exception {
+ Properties props = PropertiesUtil.deepCopy(TEST_PROPERTIES);
+ try (Connection conn = DriverManager.getConnection(getUrl(), props);) {
+ String tableName = generateUniqueName();
+ String query = "create table " + tableName + " (a_integer integer
not null "
+ + "CONSTRAINT pk PRIMARY KEY (a_integer)) SALT_BUCKETS =
256";
Review Comment:
Doesn't creating this many regions slow down the test ?
> Traversing through all the guideposts to prepare parallel scans is not
> required for salted tables when the query is point lookup
> --------------------------------------------------------------------------------------------------------------------------------
>
> Key: PHOENIX-6767
> URL: https://issues.apache.org/jira/browse/PHOENIX-6767
> Project: Phoenix
> Issue Type: Improvement
> Reporter: Rajeshbabu Chintaguntla
> Assignee: Rajeshbabu Chintaguntla
> Priority: Major
> Fix For: 5.2.0
>
>
> Currently even when the query is point lookup on salted table we are
> traversing through region boundaries or guideposts to create parallel scans
> which is not required and adding overhead to point lookup queries even we
> know the exact region or guide post need to be scanned to find the results.
> When the guide posts are high or number of regions are high leading higher
> latencies for point lookup query.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
