Issues with Synapse Workflows After Auto-Update of Self-Hosted Integration Runtime to 5.55.9315.1 (Salesforce V2 Empty-Result Parquet Schema Regression)

Question

Issues with Synapse Workflows After Auto-Update of Self-Hosted Integration Runtime to 5.55.9315.1 (Salesforce V2 Empty-Result Parquet Schema Regression)

Jeff S 5

Description

After our Self-Hosted Integration Runtime (SHIR) auto-updated to v5.55.9315.1 on November 15, 2025, Synapse pipelines that use the Salesforce V2 connector began producing Parquet files with an incomplete schema when an incremental load returns no new rows.

This did not occur on v5.52.9236.1. In a separate environment still running v5.52.9236.1, the expected behavior continues.

History / Prior Behavior (Years in Production)

For several years, across multiple connectors (including Salesforce V1, Salesforce V2, SQL Server, Db2), our pipelines have consistently produced empty Parquet files that include all expected source columns (plus one static, added control column with a file-creation timestamp) when an incremental load returns no data. This behavior remained correct with Salesforce V2 up until the SHIR auto-update on Nov 15, 2025.

Issue Details

Before (v5.52.9236.1 and earlier): When no new data matched the incremental timestamp filter, Copy Activity output a Parquet file with all source columns present (schema intact) plus our static control column (a file timestamp we add).
After (v5.55.9315.1): Under the same “no new data” condition, the Parquet file contains only the single static control column that we add. All source columns are missing from the Parquet schema, including the source column used by the incremental filter.
Scope: Observed only with the Salesforce V2 connector. Other connectors (e.g., SQL Server, Db2) still behave as expected.

Impact

Downstream processes are designed to always read the Parquet file—whether it has data or not—and rely on the full schema (with NULLs when empty). When source columns are absent, those processes fail because expected columns do not exist. While we could alter downstream logic, this appears to be a regression introduced immediately after the SHIR auto-update rather than a documented, intentional change.

Environment

Product: Synapse Pipelines / Self-Hosted Integration Runtime
Connector: Salesforce V2
Current version (failing): 5.55.9315.1
Prior version (working): 5.52.9236.1
Auto-update date: November 15, 2025

Reproduction Steps

On SHIR v5.55.9315.1, configure a Copy Activity using Salesforce V2 with an incremental timestamp filter that results in zero rows returned.
Sink is Parquet (ADLS/Blob). Add a static control column (file creation timestamp) via mapping/derived column.
Run the pipeline.
Actual: Parquet file is created but contains only the static control column; all source columns are missing.
Expected: Parquet file should be created with the full source schema plus the static control column, even when there are zero rows.

What We’ve Checked

Verified identical pipeline design and mappings on an environment running v5.52.9236.1 → works as expected (full schema present on empty results).
Confirmed other connectors (e.g., SQL Server, Db2) on v5.55.9315.1 still retain full schema on empty results.
Reviewed SHIR/Salesforce connector documentation and download pages; cannot locate a rollback package for v5.52.9236.1.

Request for Microsoft

Please confirm whether this is a known issue/bug in SHIR v5.55.9315.1 affecting Salesforce V2 when incremental loads yield zero rows.
If intentional, please share the release note/design change and recommended migration guidance.
Provide a way to download or roll back to v5.52.9236.1 (or an interim fixed build) while this is investigated.
If a hotfix or connector update is available, please advise on the build/version.

Summary

Behavior change date: Nov 15, 2025 (post-auto-update to 5.55.9315.1)
Observed only with: Salesforce V2
Actual (empty result): Parquet schema includes only our added static control column
Expected (empty result): Parquet schema includes all source columns + static control column (rows = 0)

2 answers

Your answer

Answer 1

Q&A Assist

AI generated content. This question contains an answer created with AI from Q&A Assist. Learn more

It appears that you are experiencing a regression issue with the Salesforce V2 connector in Azure Synapse Pipelines after the auto-update of the Self-Hosted Integration Runtime (SHIR) to version 5.55.9315.1. This issue results in Parquet files being generated with an incomplete schema when an incremental load returns no new rows, which is a change from the expected behavior observed in version 5.52.9236.1.

The symptoms you describe, where only the static control column is present in the output Parquet file and all source columns are missing, indicate a potential bug or unintended behavior introduced in the latest version of the SHIR. It is important to confirm whether this is a known issue affecting only the Salesforce V2 connector under the specified conditions.

For the next steps, I recommend reaching out to Microsoft support to confirm if this is a documented issue, and if so, to inquire about any available hotfixes or updates. Additionally, you may want to request guidance on rolling back to the previous version or obtaining an interim fix while this issue is being investigated.

References:

Jeff S 5 Reputation points

2025-11-17T22:06:17.43+00:00
Thanks for the reply. We need a more concrete response from Microsoft engineering rather than a general suggestion to open a ticket.

Key facts recap:

Product: Azure Synapse Pipelines – Self-Hosted Integration Runtime (SHIR)

Auto-update occurred: November 15 2025

Current SHIR version (issue present): 5.55.9315.1

Previous version (working): 5.52.9236.1

Connector: Salesforce V2 only – all other connectors continue to behave correctly.

This behavior (producing empty Parquet files with full schema when no incremental data is returned) has been stable for years across multiple connectors, including Salesforce V1, Salesforce V2 (before 11/15/2025), SQL Server, Db2.

Problem summary

When an incremental Salesforce V2 load returns zero rows:

Prior to the update, the Copy Activity created a Parquet file with all source columns present, plus one static control datetime column we add.

After the update, the Parquet file still appears, but it now contains only the static control column—all source columns are missing.

This change began immediately after SHIR auto-updated to 5.55.9315.1 and is repeatable.

Request for Microsoft clarification

Please confirm whether this is a known regression introduced in SHIR 5.55.9315.1 for the Salesforce V2 connector.

If intentional, provide a link to release notes or change documentation describing this altered behavior.

If unintentional, when can a fix or hot-patch be expected?

Is there a supported rollback path to version 5.52.9236.1 while this is investigated?

This change breaks long-established downstream processing that relies on the Parquet schema being present even when no data exists. We need to understand whether this is a confirmed defect, and if so, when to expect resolution or guidance.

Thank you—confirmation from the product group or a bug ID would be appreciated.

Answer 2

Swapnesh Panchal 1,365 Microsoft External Staff Moderator

Hi Jeff S ,
Welcome to the Microsoft Q&A and thank you for posting your questions here.

Please Upgrade SHIR to the latest build (≥ 5.57.x, current latest is 5.60.9444.1).

If upgrade isn’t possible, explicitly define schema mappings in Copy Activity or add dummy rows when zero results occur.

Downgrade to v5.52.9236.1 if schema consistency is critical.

Let us know in the comments if this solution doesn’t work for you.

Swapnesh Panchal 1,365 Reputation points Microsoft External Staff Moderator

2025-11-27T18:46:11.1866667+00:00

I hope this is helpful! Do not hesitate to let me know if you have any other questions or clarifications.
Jeff S 5 Reputation points

2025-12-01T19:23:46.3+00:00

Hi Swapnesh Panchal,

Thank you for the follow-up and the recommendation. We tested the latest SHIR version (5.60.9444.1) and confirmed that the issue still persists—Parquet files generated by the Salesforce V2 connector do not include the source schema when zero rows are returned. Our team has implemented a workaround in our process to handle this regression, but it did require additional unplanned effort. We appreciate the guidance and would like to ask if there’s a way to receive advance notice of impactful connector or SHIR behavior changes in the future to help us plan accordingly.
VRISHABHANATH PATIL 1,725 Reputation points Microsoft External Staff Moderator

2025-12-02T02:37:34.9766667+00:00
Hi @Jeff S

Thank you for your feedback. Below is a detailed explanation addressing your question.

To receive advance notice of impactful connector or SHIR behavior changes:

Subscribe to Azure Updates Use the https://azure.microsoft.com/updates/ RSS feed or email notifications for Data Factory and Synapse services. This is where connector and SHIR changes are announced.

Enable Azure Service Health Alerts Configure alerts in the Azure portal to get proactive notifications about planned maintenance or advisories affecting your services.

Monitor Release Notes Regularly Review SHIR and connector release notes before updates. If auto-update is enabled, consider setting a maintenance window and using a staging SHIR for early validation.
Jeff S 5 Reputation points

2025-12-02T13:45:25.2466667+00:00

Hi Swapnesh Panchal,

Thank you for the detailed guidance. I did review the release notes for the SHIR updates that introduced the behavior change we experienced, but I did not see any mention of this specific regression or a change in how the Salesforce V2 connector handles zero-row results. Could you please provide a link to the announcement or documentation where this change was communicated? I’d like to review it so I can better understand how such updates will be represented going forward. Thank you again for your assistance.

Share via

Issues with Synapse Workflows After Auto-Update of Self-Hosted Integration Runtime to 5.55.9315.1 (Salesforce V2 Empty-Result Parquet Schema Regression)

2 answers

Your answer