[GEODE-5061] Lucene queries can be executed on an accessor which causes the system to go into stack overflow

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: lucene
• Assignee: Unassigned
• Reporter: Nabarun Nag
• Created: 12/Apr/18 20:34
• Updated: 12/Apr/18 20:35

Description

Current:

An Apache Geode user can execute a Lucene query on an accessor without ever creating a Lucene index. This should not be allowed but it is and it causes a stack overflow when we execute a Lucene query from the accessor. This can be observed in the below test case.

 

public void asynchronousLuceneIndexCreationWithDifferentFieldsMustFail(
 RegionTestableType regionTestType) throws Exception {
 SerializableRunnableIF createIndex = () -> {
 LuceneService luceneService = LuceneServiceProvider.get(getCache());
 luceneService.createIndexFactory().addField("text").create(INDEX_NAME, REGION_NAME);
 };
 dataStore1.invoke(() -> initDataStore(createIndex, regionTestType));
 dataStore2.invoke(() -> initDataStore(createIndex, regionTestType));
 accessor.invoke(() -> initAccessor(createIndex, regionTestType));

 putDataInRegion(accessor);
 assertTrue(waitForFlushBeforeExecuteTextSearch(accessor, 60000));
 assertTrue(waitForFlushBeforeExecuteTextSearch(dataStore1, 60000));
 executeTextSearch(accessor);

 dataStore1.invoke(() -> destroyIndex());

 // re-index stored data
 AsyncInvocation ai1 = dataStore1.invokeAsync(() -> {
 recreateIndex();
 });
 
 dataStore2.invoke(() -> closeCache());
 ai1.join();
 aia.join();
 assertTrue(waitForFlushBeforeExecuteTextSearch(dataStore1, 60000));

 ai1.checkException();

 putAllDataIntoRegion(accessor);

 executeTextSearch(accessor);
}

 

However we can see that the issue is resolved when the accessor creates the Lucene Index. 

This can be seen in the below test case.

 

public void asynchronousLuceneIndexCreationWithDifferentFieldsMustFail(
 RegionTestableType regionTestType) throws Exception {
 SerializableRunnableIF createIndex = () -> {
 LuceneService luceneService = LuceneServiceProvider.get(getCache());
 luceneService.createIndexFactory().addField("text").create(INDEX_NAME, REGION_NAME);
 };
 dataStore1.invoke(() -> initDataStore(createIndex, regionTestType));
 dataStore2.invoke(() -> initDataStore(createIndex, regionTestType));
 accessor.invoke(() -> initAccessor(createIndex, regionTestType));

 putDataInRegion(accessor);
 assertTrue(waitForFlushBeforeExecuteTextSearch(accessor, 60000));
 assertTrue(waitForFlushBeforeExecuteTextSearch(dataStore1, 60000));
 executeTextSearch(accessor);

 dataStore1.invoke(() -> destroyIndex());

 // re-index stored data
 AsyncInvocation ai1 = dataStore1.invokeAsync(() -> {
 recreateIndex();
 });
 AsyncInvocation aia = accessor.invokeAsync(() -> {
 recreateIndex();
 });

 dataStore2.invoke(() -> closeCache());
 ai1.join();
 aia.join();
 assertTrue(waitForFlushBeforeExecuteTextSearch(dataStore1, 60000));

 ai1.checkException();

 putAllDataIntoRegion(accessor);

 executeTextSearch(accessor);
}

 

 

Solution:

 

Make sure that Lucene queries cannot be executed in the accessor if the accessor has not created the Lucene index used in the query.