Segfault when using DuckDBConnection concurrently.

[mediapartners]

Random segfaults while trying to run several queries asynchronously at the same time against the same connection/instance.

While I realize that's not ideal from my end and we should use a pool, but the library should have its own mechanism to prevent concurrent use or handle it correctly.

[TheRealMarcosFernando]

I'm experiencing the same issue here.

I'm using Node.js v22 in a TypeScript-based API project, and I'm trying to initialize a DuckDB instance with the code below. However, I'm consistently getting a Segmentation fault:

const instance = await DuckDBInstance.create(dbPath, {
  threads: '4',
  memory_limit: '2GB',
});

const connection = await instance.connect();

I also tried removing the custom configuration (i.e., omitting threads and memory_limit), but the segmentation fault still occurs.

[jraymakers]

@mediapartners Do you have more precise repro steps? Some details that would help:

• Are you running several asynchronous queries on the same connection or the same instance? That makes a different.
• Which API calls are you using to run the queries?
• What sort of queries are you running?
• What platform (OS & arch) are you running on?

[jraymakers]

@TheRealMarcosFernando The original issue is about running multiple asynchronous queries at the same time. The issue you report is different - you're just creating the instance & connection, not running any queries.

Since pretty much every use of the library requires creating an instance and a connection and no one else has reported this, I suspect something specific to your environment is related.

Please open a separate issue with more details about your environment, such as:

• What is dbPath?
• What OS & arch?
• How are you running the Node script?
• What else might be happening on the machine at the same time?

[mediapartners]

@jraymakers DuckDBConnection.runAndReadAll and manually creating a prepared statement.

The fix for me was to create a new connection for every query and do closeSync afterwards.

I tried creating a code snippet that ran a bunch of queries concurrently, while I was able to replicate a few crashes it wasn't consistent. Also its possible the motherduck connection and parameter binding are required to replicate this problem

[elefeint]

Could you try again with v1.3.2? There was a duckdb Windows issue that got fixed in 1.3.2

[mediapartners]

@elefeint thanks, but I am on linux

[jraymakers]

@hrl20 tracked this down and found a fix, which has been merged. I'm working on a getting a release out shortly.

[jraymakers]

Fix released in 1.3.2-alpha.25.