[BUG] <title>NPE when SETTING_NUMBER_OF_REPLICAS is null

[jaytiBansal]

Describe the bug

in OS 2.15, IndexSetting.java, method getNumberOfReplicas will give null pointer exception in scenarios when null is returned from settings.getAsInt methods because the return type of this method is int (not Integer).

public int getNumberOfReplicas() {
return settings.getAsInt(IndexMetadata.SETTING_NUMBER_OF_REPLICAS, null);
}

Related component

Indexing

To Reproduce

Create an index with replica count passed as null.

Expected behavior

It should treat the default value as 1 for replica count is not passed.

Additional Details

[2024-07-16T17:02:27,740][DEBUG][o.o.c.m.MetadataCreateIndexService] [local_10-10] [ds7_dfs_] failed to create
java.lang.NullPointerException: Cannot invoke "java.lang.Integer.intValue()" because the return value of "org.opensearch.common.settings.Settings.getAsInt(String, java.lang.Integer)" is null
at org.opensearch.index.IndexSettings.getNumberOfReplicas(IndexSettings.java:1248)
at org.opensearch.indices.IndicesService.createIndexService(IndicesService.java:867)
at org.opensearch.indices.IndicesService.withTempIndexService(IndicesService.java:830)
at org.opensearch.cluster.metadata.MetadataCreateIndexService.applyCreateIndexWithTemporaryService(MetadataCreateIndexService.java:502)
at org.opensearch.cluster.metadata.MetadataCreateIndexService.applyCreateIndexRequestWithV1Templates(MetadataCreateIndexService.java:643)
at org.opensearch.cluster.metadata.MetadataCreateIndexService.applyCreateIndexRequest(MetadataCreateIndexService.java:464) ~
at org.opensearch.cluster.metadata.MetadataCreateIndexService.applyCreateIndexRequest(MetadataCreateIndexService.java:471)
at org.opensearch.cluster.metadata.MetadataCreateIndexService$1.execute(MetadataCreateIndexService.java:377)
at org.opensearch.cluster.ClusterStateUpdateTask.execute(ClusterStateUpdateTask.java:67)
at org.opensearch.cluster.service.MasterService.executeTasks(MasterService.java:904)
at org.opensearch.cluster.service.MasterService.calculateTaskOutputs(MasterService.java:456)
at org.opensearch.cluster.service.MasterService.runTasks(MasterService.java:316)
at org.opensearch.cluster.service.MasterService$Batcher.run(MasterService.java:227)
at org.opensearch.cluster.service.TaskBatcher.runIfNotProcessed(TaskBatcher.java:204)
at org.opensearch.cluster.service.TaskBatcher$BatchedTask.run(TaskBatcher.java:242)
at org.opensearch.common.util.concurrent.ThreadContext$ContextPreservingRunnable.run(ThreadContext.java:882)
at org.opensearch.common.util.concurrent.PrioritizedOpenSearchThreadPoolExecutor$TieBreakingPrioritizedRunnable.runAndClean(PrioritizedOpenSearchThreadPoolExecutor.java:283)

[chishui]

The default value of replica count is not always 1, user can configure the default value through this cluster setting cluster.default_number_of_replicas, so simply changing default value from null to 1 won't work for every scenario.

I think it's better to treat index.number_of_replicas = null the same way as when index.number_of_replicas is missing from the payload.

[chishui]

Fixed in #14812

[mgodwan]

[Indexing Triage Meeting 07/22]

@chishui Thanks for working on this. I've assigned the issue to you. Please let us know if you are planning to continue working on it, and targeting to merge it within a release window.

[chishui]

@mgodwan I'm working on the PR #14812 and waiting for this to be approve and merged

[sandeshkr419]

Complimenting my comment on #14810, if we have an agreement that null should be treated as 0 - then the same should apply here to keep consistency.