[BUG] Unable to enable remote state repository independently of data repositories

[BhumikaSaini-Amazon]

Describe the bug

Enabling just remote state repo (support added via PR #11858 ) starts a remote store migration. This migration doesn’t go through. The shards stay unassigned.

Related component

Storage

To Reproduce

1. Launch a docrep cluster with remote state enabled. Do not enable remote segment and remote translog repositories.
2. Create an index.
3. Shard creation fails and shards stay unassigned.

Expected behavior

Enabling only remote state repository should not start a migration to remote store

Additional Details

Exception stack trace

[2024-05-02T13:17:25,436][INFO ][o.o.i.IndexService       ] [node-1] [idx1] DocRep shard [idx1][3] is migrating to remote
[2024-05-02T13:17:25,436][WARN ][o.o.i.c.IndicesClusterStateService] [node-1] [idx1][3] marking and sending shard failed due to [failed to create shard]
java.lang.NullPointerException: Cannot invoke "Object.hashCode()" because "key" is null
    at java.base/java.util.concurrent.ConcurrentHashMap.get(ConcurrentHashMap.java:936) ~[?:?]
    at org.opensearch.repositories.RepositoriesService.repository(RepositoriesService.java:568) ~[opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.index.store.RemoteSegmentStoreDirectoryFactory.newDirectory(RemoteSegmentStoreDirectoryFactory.java:61) ~[opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.index.IndexService.createShard(IndexService.java:512) ~[opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.indices.IndicesService.createShard(IndicesService.java:1025) ~[opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.indices.IndicesService.createShard(IndicesService.java:213) ~[opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.indices.cluster.IndicesClusterStateService.createShard(IndicesClusterStateService.java:672) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.indices.cluster.IndicesClusterStateService.createOrUpdateShards(IndicesClusterStateService.java:649) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.indices.cluster.IndicesClusterStateService.applyClusterState(IndicesClusterStateService.java:294) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.cluster.service.ClusterApplierService.callClusterStateAppliers(ClusterApplierService.java:608) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.cluster.service.ClusterApplierService.callClusterStateAppliers(ClusterApplierService.java:595) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.cluster.service.ClusterApplierService.applyChanges(ClusterApplierService.java:563) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.cluster.service.ClusterApplierService.runTask(ClusterApplierService.java:486) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.cluster.service.ClusterApplierService$UpdateTask.run(ClusterApplierService.java:188) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.common.util.concurrent.ThreadContext$ContextPreservingRunnable.run(ThreadContext.java:854) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.common.util.concurrent.PrioritizedOpenSearchThreadPoolExecutor$TieBreakingPrioritizedRunnable.runAndClean(PrioritizedOpenSearchThreadPoolExecutor.java:283) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at org.opensearch.common.util.concurrent.PrioritizedOpenSearchThreadPoolExecutor$TieBreakingPrioritizedRunnable.run(PrioritizedOpenSearchThreadPoolExecutor.java:246) [opensearch-3.0.0-SNAPSHOT.jar:3.0.0-SNAPSHOT]
    at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144) [?:?]
    at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642) [?:?]
    at java.base/java.lang.Thread.run(Thread.java:1583) [?:?]

Proposed solution
The migration flow uses the isRemoteStoreNode( ) method. This method checks for the presence of any remote_store node attribute on the node:

OpenSearch/server/src/main/java/org/opensearch/cluster/node/DiscoveryNode.java

Lines 464 to 471 in ed33488

	/**
	* Returns whether the node is a remote store node.
	*
	* @return true if the node contains remote store node attributes, false otherwise
	*/
	public boolean isRemoteStoreNode() {
	return this.getAttributes().keySet().stream().anyMatch(key -> key.startsWith(REMOTE_STORE_NODE_ATTRIBUTE_KEY_PREFIX));
	}

Given that the state repo should be independent now, we should have distinct methods to identify whether the cluster state or data is remote-backed.

[sulthan309]

Hi, I would like to contribute to resolve this bug. Can you please assign this to me?

[BhumikaSaini-Amazon]

Thank you @sulthan309 for volunteering!

We are tracking this bugfix for the 2.15 release. We want to get the fix merged to main and backported to 2.x by the code freeze date of 10th June (calendar).

Please do check how the method is used at various places. That will help with identifying the changes we need. If you need more info anytime, please let us know.

Looking forward to your contribution!

[sulthan309]

Thank you for assigning this ticket to me.

@BhumikaSaini-Amazon Sure i will go through the code and reach out if needed.

[rookuu]

We're running into this bug also in 2.15. Do we have a new target release?