[BUG] java.lang.NullPointerException: Cannot invoke "com.azure.cosmos.models.CosmosBulkItemResponse.getCosmosDiagnostics()" because the return value of "com.azure.cosmos.models.CosmosBulkOperationResponse.getResponse()" is null

[petercooperjr-spls]

Describe the bug
On Aug. 12 around 19:20 UTC, one of our services (running in AKS) seemed to have a connection issue getting to our CosmosDB instance. We got errors while calling .saveAll() on a Spring Data Cosmos Repository on a collection of entities three times. While the application recovered the connection and later requests worked, there was an NullPointerException logged that at the very least made things a little confusing to figure out, and I think that without it, it might have automatically retried those specific data saves.

Exception or Stack Trace

2026-08-12T19:20:20.709Z WARN 1 --- [dispatch] [rntbd-epoll-3-4] .c.i.d.r.RntbdClientChannelHealthChecker : [id: 0x9734fd0f, L:/100.96.59.236:33428 - R:cdb-ms-prod-eastus2-be122.documents.azure.com/20.15.13.186:14142] health check failed due to transit timeout on write threshold hit: [rntbdContext: Optional[RntbdContext({"activityId":"3715933b-5ff5-4f87-910f-4ccb2ca5cc66","clientVersion":"","idleTimeoutInSeconds":120,"protocolVersion":0,"serverProperties":{"agent":"DocumentDB Server","version":" version=2.14.0.0"},"unauthenticatedTimeoutInSeconds":25,"statusCode":200})],lastChannelRead: 2026-08-12T19:20:04.312758445Z, transitTimeoutWriteCount: 1, timeoutOnWriteThreshold: 1, timeoutOnWriteTimeLimitInNanos: 6,000,000,000]clientVmId: uuid:acfda9ad-b196-42d3-b679-aab15d30aeec, clientUsedMemory: 644018 KB, clientAvailableMemory: 1453134 KB, clientSystemCpuLoad: (2026-08-12T19:19:52.045584431Z 21.5%), (2026-08-12T19:19:57.045598301Z 35.4%), (2026-08-12T19:20:02.045592776Z 30.2%), (2026-08-12T19:20:07.045591576Z 23.9%), (2026-08-12T19:20:12.045594423Z 15.4%), (2026-08-12T19:20:17.045587114Z 25.6%), clientAvailableProcessors: 2]
2026-08-12T19:20:20.710Z WARN 1 --- [dispatch] [rntbd-epoll-3-4] c.a.c.i.d.GoneAndRetryWithRetryPolicy : Operation will NOT be retried. Write operations which failed due to transient transport errors can not be retried safely when sending the request to the service because they aren't idempotent. Current attempt 1, Exception:
{"ClassName":"GoneException","userAgent":"azsdk-java-cosmos/4.79.0 Linux/5.15.0-1111-azure JRE/17.0.20","statusCode":410,"resourceAddress":"rntbd://cdb-ms-prod-eastus2-be122.documents.azure.com:14142/apps/96861dd2-ea69-4273-bb2c-446d5fadc16e/services/c6d579b9-537f-44a6-9978-2b85226bf347/partitions/9afedafd-54cf-4263-80e6-3187ce0c01c1/replicas/134243081869070243p/","innerErrorMessage":"ChannelHandlerContext(RntbdRequestManager#0, [id: 0x9734fd0f, L:/100.96.59.236:33428 - R:cdb-ms-prod-eastus2-be122.documents.azure.com/20.15.13.186:14142]) closed exceptionally with 1 pending requests","causeInfo":"[class: class com.azure.cosmos.implementation.directconnectivity.rntbd.RntbdRequestManager$UnhealthyChannelException, message: health check failed, reason: [id: 0x9734fd0f, L:/100.96.59.236:33428 - R:cdb-ms-prod-eastus2-be122.documents.azure.com/20.15.13.186:14142] health check failed due to transit timeout on write threshold hit: [rntbdContext: Optional[RntbdContext({\"activityId\":\"3715933b-5ff5-4f87-910f-4ccb2ca5cc66\",\"clientVersion\":\"\",\"idleTimeoutInSeconds\":120,\"protocolVersion\":0,\"serverProperties\":{\"agent\":\"DocumentDB Server\",\"version\":\" version=2.14.0.0\"},\"unauthenticatedTimeoutInSeconds\":25,\"statusCode\":200})],lastChannelRead: 2026-08-12T19:20:04.312758445Z, transitTimeoutWriteCount: 1, timeoutOnWriteThreshold: 1, timeoutOnWriteTimeLimitInNanos: 6,000,000,000]clientVmId: uuid:acfda9ad-b196-42d3-b679-aab15d30aeec, clientUsedMemory: 644018 KB, clientAvailableMemory: 1453134 KB, clientSystemCpuLoad: (2026-08-12T19:19:52.045584431Z 21.5%), (2026-08-12T19:19:57.045598301Z 35.4%), (2026-08-12T19:20:02.045592776Z 30.2%), (2026-08-12T19:20:07.045591576Z 23.9%), (2026-08-12T19:20:12.045594423Z 15.4%), (2026-08-12T19:20:17.045587114Z 25.6%), clientAvailableProcessors: 2]]","responseHeaders":"{x-ms-substatus=0}","requestHeaders":"[x-ms-cosmos-batch-atomic=false, x-ms-cosmos-sdk-supportedcapabilities=1, Accept=application/json, x-ms-cosmos-is-batch-request=true, x-ms-date=Wed, 12 Aug 2026 19:20:19 GMT, x-ms-documentdb-collection-rid=Ha42AP+cbB4=, x-ms-documentdb-partitionkeyrangeid=0, x-ms-cosmos-batch-continue-on-error=true, x-ms-client-retry-attempt-count=0, x-ms-remaining-time-in-ms-on-client=60000, Content-Type=application/json]"}
at com.azure.cosmos.implementation.directconnectivity.rntbd.RntbdRequestManager.completeAllPendingRequestsExceptionally(RntbdRequestManager.java:948) ~[azure-cosmos-4.79.0.jar!/:4.79.0]
at com.azure.cosmos.implementation.directconnectivity.rntbd.RntbdRequestManager.exceptionCaught(RntbdRequestManager.java:343) ~[azure-cosmos-4.79.0.jar!/:4.79.0]
at com.azure.cosmos.implementation.directconnectivity.rntbd.RntbdRequestManager.lambda$userEventTriggered$0(RntbdRequestManager.java:383) ~[azure-cosmos-4.79.0.jar!/:4.79.0]
at io.netty.util.concurrent.DefaultPromise.notifyListener0(DefaultPromise.java:604) ~[netty-common-4.1.135.Final.jar!/:4.1.135.Final]

…

Caused by: com.azure.cosmos.implementation.directconnectivity.rntbd.RntbdRequestManager$UnhealthyChannelException: health check failed, reason: [id: 0x9734fd0f, L:/100.96.59.236:33428 - R:cdb-ms-prod-eastus2-be122.documents.azure.com/20.15.13.186:14142] health check failed due to transit timeout on write threshold hit: [rntbdContext: Optional[RntbdContext({"activityId":"3715933b-5ff5-4f87-910f-4ccb2ca5cc66","clientVersion":"","idleTimeoutInSeconds":120,"protocolVersion":0,"serverProperties":{"agent":"DocumentDB Server","version":" version=2.14.0.0"},"unauthenticatedTimeoutInSeconds":25,"statusCode":200})],lastChannelRead: 2026-08-12T19:20:04.312758445Z, transitTimeoutWriteCount: 1, timeoutOnWriteThreshold: 1, timeoutOnWriteTimeLimitInNanos: 6,000,000,000]clientVmId: uuid:acfda9ad-b196-42d3-b679-aab15d30aeec, clientUsedMemory: 644018 KB, clientAvailableMemory: 1453134 KB, clientSystemCpuLoad: (2026-08-12T19:19:52.045584431Z 21.5%), (2026-08-12T19:19:57.045598301Z 35.4%), (2026-08-12T19:20:02.045592776Z 30.2%), (2026-08-12T19:20:07.045591576Z 23.9%), (2026-08-12T19:20:12.045594423Z 15.4%), (2026-08-12T19:20:17.045587114Z 25.6%), clientAvailableProcessors: 2]
2026-08-12T19:20:20.710Z WARN 1 --- [dispatch] [rntbd-epoll-3-2] .c.i.d.r.RntbdClientChannelHealthChecker : [id: 0xee099cb0, L:/100.96.59.236:41392 - R:cdb-ms-prod-eastus2-be122.documents.azure.com/20.15.13.186:14142] health check failed due to transit timeout on write threshold hit: [rntbdContext: Optional[RntbdContext({"activityId":"d1d308a5-d71d-415e-a645-6d0534c663bd","clientVersion":"","idleTimeoutInSeconds":120,"protocolVersion":0,"serverProperties":{"agent":"DocumentDB Server","version":" version=2.14.0.0"},"unauthenticatedTimeoutInSeconds":25,"statusCode":200})],lastChannelRead: 2026-08-12T19:20:04.312309160Z, transitTimeoutWriteCount: 1, timeoutOnWriteThreshold: 1, timeoutOnWriteTimeLimitInNanos: 6,000,000,000]clientVmId: uuid:acfda9ad-b196-42d3-b679-aab15d30aeec, clientUsedMemory: 644018 KB, clientAvailableMemory: 1453134 KB, clientSystemCpuLoad: (2026-08-12T19:19:52.045584431Z 21.5%), (2026-08-12T19:19:57.045598301Z 35.4%), (2026-08-12T19:20:02.045592776Z 30.2%), (2026-08-12T19:20:07.045591576Z 23.9%), (2026-08-12T19:20:12.045594423Z 15.4%), (2026-08-12T19:20:17.045587114Z 25.6%), clientAvailableProcessors: 2]
2026-08-12T19:20:20.713Z INFO 1 --- [dispatch] [smos-parallel-5] c.a.c.implementation.batch.BulkExecutor : BulkExecutor.execute flux terminated - Signal: cancel - # left items 1, Context: BulkExecutor-38657[n/a], Thread[Name: spring-data-cosmos-parallel-5,Group: main, isDaemon: true, Id: 123]

…

2026-08-12T19:20:20.720Z ERROR 1 --- [dispatch] [smos-parallel-5] a.w.r.e.AbstractErrorWebExceptionHandler : [5245435c-2417817] 500 Server Error for HTTP POST "/api/external/routePlans/252430/2026-08-13/ddb55c56-1721-49ca-a907-59b5c683fb1a/moveStops"
java.lang.NullPointerException: Cannot invoke "com.azure.cosmos.models.CosmosBulkItemResponse.getCosmosDiagnostics()" because the return value of "com.azure.cosmos.models.CosmosBulkOperationResponse.getResponse()" is null
at com.azure.spring.data.cosmos.core.ReactiveCosmosTemplate.lambda$insertAll$2(ReactiveCosmosTemplate.java:557) ~[azure-spring-data-cosmos-6.3.0.jar!/:6.3.0]

To Reproduce

That certainly is the tricky part; I don't know how to reproduce a sporadic connection issue with Cosmos.

Code Snippet
While there's plenty of things before and after, the line that had the problem was

stopRepository.saveAll(newStops).collectList()

Expected behavior
No NullPointerException being logged while it worked on handling the temporary connection problem.

Setup (please complete the following information):

• OS: AKS, using Docker image mcr.microsoft.com/openjdk/jdk:17-distroless, underlying Linux kernel 5.15.0-1111-azure
• IDE: IntelliJ IDEA Ultimate 2026.1.4
• Library/Libraries: azure-spring-data-cosmos 6.3.0
• App Server/Environment: AKS
• Frameworks: Spring Boot 3.5.16

Additional context
I think that it's trying to log information about the request, but since the request failed there wasn't actually an object to get data about.

Information Checklist
Kindly make sure that you have added all the following information above and checkoff the required fields otherwise we will treat the issuer as an incomplete report

• Bug Description Added
• Repro Steps Added
• Setup information Added

[github-actions]

Thank you for your feedback. Tagging and routing to the team member(s) best able to assist.

//cc: Kushagra Thapar (@kushagraThapar) pjohari-ms Theo van Kraay (@TheovanKraay)

[github-actions]

🎯 Agentic Issue Triage

Summary: A NullPointerException occurs in ReactiveCosmosTemplate when CosmosBulkOperationResponse.getResponse() returns null after a transient RNTBD connection failure, causing an unhandled NPE during bulk write diagnostics logging in azure-spring-data-cosmos 6.3.0.

📋 Issue Details

Package: com.azure:azure-spring-data-cosmos:6.3.0 (also involves com.azure:azure-cosmos:4.79.0)
Affected API: ReactiveCosmosTemplate.lambda$insertAll$2 (line 557) — CosmosBulkOperationResponse.getResponse().getCosmosDiagnostics()
Scenarios:

• Transient RNTBD channel health check failure (transit timeout on write threshold) causes a GoneException on a batch write
• Spring Data Cosmos saveAll() calls into BulkExecutor, which receives a CosmosBulkOperationResponse with a null response field
• ReactiveCosmosTemplate dereferences getResponse() without a null check, causing an NPE

Root ask: Null check added in ReactiveCosmosTemplate so that transient Cosmos connectivity issues do not also produce an NPE log; ideally the retry logic should also handle this case automatically.

🔎 Debugging / Reproduction Notes

The NPE originates in ReactiveCosmosTemplate.java:557 where getCosmosDiagnostics() is called on the result of getResponse(), which is null when the underlying RNTBD channel was closed due to health-check failure. The GoneAndRetryWithRetryPolicy explicitly logs that the write operation will NOT be retried (non-idempotent write). The BulkExecutor flux is then cancelled with 1 item remaining, leaving a CosmosBulkOperationResponse with a null response.

Suggested investigation steps:

1. Review ReactiveCosmosTemplate.java around line 557 — add a null-check on getResponse() before calling getCosmosDiagnostics()
2. Determine whether a CosmosBulkOperationResponse with a null response should result in the exception being re-thrown or swallowed gracefully
3. Check whether the non-retry policy for write GoneExceptions is intentional in Spring Data Cosmos batch path and whether a fallback or user-visible error is more appropriate

🏷️ Label Confidence

• Category: Client — the NPE is in SDK library code (ReactiveCosmosTemplate), not a REST API or service-side behavior issue
• Service: Cosmos — the package azure-spring-data-cosmos lives under sdk/cosmos/ and involves the Azure Cosmos DB SDK
• Confidence: High — the package, service directory, and failing code path are explicitly named in the stack trace

👥 Owner Routing

• Matched CODEOWNERS entry: # ServiceLabel: %Cosmos (line 292) — "Cosmos" matches the predicted service label
• AzureSdkOwners: none listed
• ServiceOwners: @kushagraThapar @pjohari-ms @TheovanKraay
• Routing action: Added "Service Attention" and "needs-team-attention" labels; @mentioned ServiceOwners
• Scan notes: Scanned bottom-to-top; %Azure.Spring - Cosmos (line 549) requires label "Azure.Spring - Cosmos" which is not predicted → skipped; %Cosmos (line 292) matched on predicted "Cosmos" label → stop

Generated by Agentic Triage for #50148 · 37.4 AIC · ⌖ 5.67 AIC · ⊞ 11K · ◷