Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix record skipping when querying paginated data across shards #3061

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -707,6 +707,13 @@ public enum Index {
"calcs",
getMappingFile("calcs_index_mappings.json"),
"src/test/resources/calcs.json"),
// Calcs has enough records for shards to be interesting, but updating the existing mapping with shards in-place
// breaks existing tests. Aside from introducing a primary shard setting > 1, this index is identical to CALCS.
CALCS_WITH_SHARDS(
TestsConstants.TEST_INDEX_CALCS,
"calcs",
getMappingFile("calcs_with_shards_index_mappings.json"),
"src/test/resources/calcs.json"),
DATE_FORMATS(
TestsConstants.TEST_INDEX_DATE_FORMATS,
"date_formats",
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -5,19 +5,23 @@

package org.opensearch.sql.sql;

import static org.opensearch.sql.legacy.TestsConstants.TEST_INDEX_PHRASE;

import java.io.IOException;
import java.util.ArrayList;
import java.util.List;

import org.json.JSONObject;
import org.junit.After;
import org.junit.Test;
import org.opensearch.client.ResponseException;
import org.opensearch.sql.legacy.SQLIntegTestCase;

import static org.opensearch.sql.legacy.TestsConstants.*;

public class PaginationWindowIT extends SQLIntegTestCase {
@Override
public void init() throws IOException {
loadIndex(Index.PHRASE);
loadIndex(Index.CALCS_WITH_SHARDS);
}

@After
Expand Down Expand Up @@ -92,4 +96,39 @@ public void testFetchSizeLargerThanResultWindowFails() throws IOException {
() -> executeQueryTemplate("SELECT * FROM %s", TEST_INDEX_PHRASE, window + 1));
resetMaxResultWindow(TEST_INDEX_PHRASE);
}

@Test
public void testMultiShardPagesEqualsActualData() throws IOException {
// A bug made it so when pulling unordered data from an index with multiple shards, data gets lost if the fetchSize
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

pulling unordered data from an index
it is not accurate, it is ordered data (sord by _doc)?

// is not a multiple of the shard count. This tests that, for an index with 4 shards, pulling one page of 10 records
// is equivalent to pulling two pages of 5 records.

var query = "SELECT key from " + TEST_INDEX_CALCS;

var expectedResponse = new JSONObject(executeFetchQuery(query, 10, "jdbc"));
var expectedRows = expectedResponse.getJSONArray("datarows");

List<String> expectedKeys = new ArrayList<>();
for (int i = 0; i < expectedRows.length(); i++) {
expectedKeys.add(expectedRows.getJSONArray(i).getString(0));
}

var actualPage1 = new JSONObject(executeFetchQuery(query, 5, "jdbc"));

var actualRows1 = actualPage1.getJSONArray("datarows");
var cursor = actualPage1.getString("cursor");
var actualPage2 = executeCursorQuery(cursor);

var actualRows2 = actualPage2.getJSONArray("datarows");

List<String> actualKeys = new ArrayList<>();
for (int i = 0; i < actualRows1.length(); i++) {
actualKeys.add(actualRows1.getJSONArray(i).getString(0));
}
for (int i = 0; i < actualRows2.length(); i++) {
actualKeys.add(actualRows2.getJSONArray(i).getString(0));
}

assertEquals(expectedKeys, actualKeys);
}
}
Original file line number Diff line number Diff line change
@@ -0,0 +1,99 @@
{
"mappings" : {
"properties" : {
"key" : {
"type" : "keyword"
},
"num0" : {
"type" : "double"
},
"num1" : {
"type" : "double"
},
"num2" : {
"type" : "double"
},
"num3" : {
"type" : "double"
},
"num4" : {
"type" : "double"
},
"str0" : {
"type" : "keyword"
},
"str1" : {
"type" : "keyword"
},
"str2" : {
"type" : "keyword"
},
"str3" : {
"type" : "keyword"
},
"int0" : {
"type" : "integer"
},
"int1" : {
"type" : "integer"
},
"int2" : {
"type" : "integer"
},
"int3" : {
"type" : "integer"
},
"bool0" : {
"type" : "boolean"
},
"bool1" : {
"type" : "boolean"
},
"bool2" : {
"type" : "boolean"
},
"bool3" : {
"type" : "boolean"
},
"date0" : {
"type" : "date",
"format": "year_month_day"
},
"date1" : {
"type" : "date",
"format": "year_month_day"
},
"date2" : {
"type" : "date",
"format": "year_month_day"
},
"date3" : {
"type" : "date",
"format": "year_month_day"
},
"time0" : {
"type" : "date",
"format": "date_time_no_millis"
},
"time1" : {
"type" : "date",
"format": "hour_minute_second"
},
"datetime0" : {
"type" : "date",
"format": "date_time_no_millis"
},
"datetime1" : {
"type" : "date"
},
"zzz" : {
"type" : "keyword"
}
}
},
"settings": {
"index": {
"number_of_shards": 4
}
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -189,6 +189,9 @@ public OpenSearchResponse searchWithPIT(Function<SearchRequest, SearchResponse>
// Set sort field for search_after
if (this.sourceBuilder.sorts() == null) {
this.sourceBuilder.sort(DOC_FIELD_NAME, ASC);
// Workaround to preserve sort location more exactly,
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if this is workaround, could u add the long-term solution issue?

// see https://github.com/opensearch-project/sql/pull/3061
this.sourceBuilder.sort("_id", ASC);
Copy link
Contributor

@fddattal fddattal Oct 9, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see there are a couple places in the code where we sort on doc field.
Can you help me understand why we only need this here?

https://github.com/search?q=repo%3Aopensearch-project%2Fsql+%22DOC_FIELD_NAME%22&type=code

}
SearchRequest searchRequest = new SearchRequest().source(this.sourceBuilder);
this.searchResponse = searchAction.apply(searchRequest);
Expand Down
Loading