ConstraintValidationFailed - When importing MSSQLHound collect over an exitsant collection with Sharphound

[Blatzy]

Hi !
I might found an issue when importing an MSSQLHound collection in Bloodhound CE. Here is a quick recap of the issue (with Claude Opus 4.5).

Description

When importing MSSQLHound data into a BloodHound database that already contains SharpHound data, the ingest fails with a constraint violation error.

The issue occurs because the MERGE statement creates a new node with label MSSQL_Base and then tries to add the Computer label, which violates the uniqueness constraint on Computer.objectid when the computer was already collected by SharpHound.

MSSQHound and Sharphound were used from a non-domain-joined machine.

Error message

time=2026-02-03T11:13:26.235Z level=ERROR message="Ingest task failed" task_id=2 file=mssql-bloodhound-20260127-180831.zip err="update nodes by error on statement (unwind $p as p merge (n:MSSQL_Base ) set n += p, n:MSSQL_Base, n:Computer, n:Base;): Neo4jError: Neo.ClientError.Schema.ConstraintValidationFailed (Node(99) already exists with label Computer and property objectid = 'S-1-5-21-REDACTED-147982')"

Steps to reproduce

1. Run SharpHound collection from a non-domain-joined machine
2. Ingest SharpHound data into BloodHound
3. Run MSSQLHound collection from a non-domain joined machine
4. Attempt to ingest MSSQLHound data into BloodHound

Expected behavior

MSSQLHound should merge with existing Computer nodes rather than attempting to create new ones. The MSSQL properties and labels should be added to the existing node.

Suggested fix

The query should MERGE on the existing Computer node instead of creating a new one:

MERGE (n:Computer {objectid:p.objectid})
SET n += p, n:MSSQL_Base, n:Base

Environment

• BloodHound CE version: v8.5.2
• Collection order: SharpHound first, then MSSQLHound
• Both collections run from non-domain joined machines

Best regards,

[Mayyhem]

Hey @Blatzy! Thank you for taking the time to report this issue!

Are you using neo4j for your backend database? I wasn't able to reproduce this issue when using postgres.

You can switch by opening your docker-compose file and setting bhe_graph_driver: ${bhe_graph_driver:-pg}.

I will also break up Base objects into their own output file in a future version to make this work in neo4j. In the meantime, for a workaound in neo4j, you could try removing this from the top of the MSSQLHound JSON output before uploading:

  "metadata": {
    "source_kind": "MSSQL_Base"
  },

However, please note that this will make it more difficult to bulk delete the data from MSSQLHound in the Database Management page since the uploaded objects won't be associated with the MSSQL_Base source kind.