Server Error. Message : The metadata for the Delta table 'table_name' with the version 'n' could not

[abhilash_o_p]

Hi All,

I am facing this issue on a shortcut created from ADLSG2.

Error:

Server Error.
Message : The metadata for the Delta table '<TableName>' with the version '46' could not be found.
Error Code : MissingMetadataForDeltaTableVersion
Error source : 1

Error

I checked the _delta_log folder & could see 00000000000000000046.json & 00000000000000000046.crc files. The Json file has only 'commitInfo'. There is no 'add' or 'remove' present. The shortcut table is universally used by other pipelines so recreating the shortcut is not an option (I tried recreating the shortcut and it resolves the issue). Kindly help on how I can resolve this?

[v-karpurapud]

Hi @abhilash_o_p 

We have not yet heard back from you about whether our response addressed your query. If it did not, please share more details so we can assist you more effectively.

Thank You.

[abhilash_o_p]

Hi @v-karpurapud

It is not possible for us to make changes in the source data due to limited access. What we have done is to delete and recreate the shortcuts for now. Most of the time that also won't help and will get this error "[{"message":"\"Msg 24596, Level 16, State 31, Line 1\""}]" from SQL Endpoint. The same data is used in oru databricks pipeline as well which is the primary and there is no issue there for the same tables.

We have done this as a part of PoC to see the potential and possibilities in Fabric. We expect a fix from Fabric for this issue as the source table can change on a daily basis & the same issue in Fabric is not coming in Trino which connects to the Databricks metastore. We are about to open a ticket for this.

Thank you.

[v-karpurapud]

Hi @abhilash_o_p 

Thank you for the update. We understand you have limited access to the source data and cannot modify the Delta table directly. As noted, recreating the shortcut is only a temporary workaround and does not always resolve the problem.
This issue arises because Microsoft Fabric shortcuts always point to the latest Delta log version. Currently, Fabric does not support reading historical versions or handling incomplete Delta commits as other platforms like Databricks or Trino do, especially when using the Databricks metastore. Fabric’s shortcut approach is stricter in interpreting Delta metadata.

We suggest submitting a support ticket to Microsoft so the product team can review your situation and consider it for resolution or future improvements.
 

You can submit a ticket through the Microsoft Power BI Support Portal:

How to create a Fabric and Power BI Support ticket - Power BI | Microsoft Learn

Regards,

Karpurapu D,

Microsoft Fabric Community Support Team.

[abhilash_o_p]

@v-karpurapud - Thank you for your help. I have submitted a ticket for the same.

[v-karpurapud]

Hi @abhilash_o_p 

Could you let us know if the issue was resolved after creating the support case? 

Thank You.

[abhilash_o_p]

Hi @v-karpurapud - This is still not resolved. No update on support case as well. I believe it might take some time. Will update here when I get a update from the Fabric support team.

[v-karpurapud]

Hi @abhilash_o_p 

Thank you for the updated. Looking forward to your response.

Regards,
Karpurapu D.

[v-karpurapud]

Hi ,

Thank you for keeping us informed. I hope your issue gets resolved soon. If it does, please update us here.

Best Regards,
Karpurapu D.

[v-karpurapud]

Hi @abhilash_o_p 

 Apologize for the earlier misunderstanding regarding thereading a previous version of the Delta table through a Fabric shortcut. I want to clarify that while Spark notebooks allow Delta Lake time travel using options like versionAsOf, Microsoft Fabric shortcuts do not currently support this feature. Fabric shortcuts always point to the latest version of the Delta log and do not let you access or specify previous versions. So, using versionAsOf to read an earlier version (like version 45) is only possible in Spark environments, not with Fabric shortcuts.

To address this limitation, it is recommended to trigger a new valid Delta version by making a minor write (such as a dummy update or insert). This will create a new version (for example, version 47) with complete metadata, enabling the shortcut to reference a valid state and work properly without needing to recreate it.


Thank You.

[v-karpurapud]

Hi @abhilash_o_p 

Thank you for contacting the Microsoft Fabric Community Forum.

The error “The metadata for the Delta table with the version '46' could not be found” occurs because the Delta log entry for version 46 is incomplete. Although the _delta_log folder contains the expected JSON and CRC files, the JSON file only includes a commitInfo section and lacks the necessary add or remove actions that record data changes. This typically results from transactions that commit without modifying any data, such as schema-only changes or no-op writes. While Delta Lake supports these commits, certain engines like Microsoft Fabric when using ADLS Gen2 shortcuts require each version to represent a valid table state.

Since Fabric shortcuts read the Delta log directly, they cannot interpret version 46 correctly, leading to the MissingMetadataForDeltaTableVersion error. Recreating the shortcut temporarily resolves the issue by refreshing the metadata, but the root cause remains.

To address this, consider performing a minor write operation (such as a dummy update or inserting and deleting a row) to generate a new version with the correct metadata.Alternatively, reading the table at an earlier version (e.g., version 45) using versionAsOf can serve as a temporary workaround. To prevent similar issues, ensure all Delta write operations include a meaningful change so that each log version accurately reflects the table’s state and can be properly read by tools like Fabric.

Regards,

Karpurapu D,

Microsoft Fabric Community Support Team.

[abhilash_o_p]

@v-karpurapud - Thank you for your response.

How can we read the previous version of the table shortcut? When we check the table metadata (expanding to see the columns) this error is thrown. When we read from spark notebook, it seems its not any problem.