[web] Random "SqliteException(10): disk I/O error" when using Drift across multiple Chrome tabs

[Silko5]

Hi,
I’m running into a concurrency issue on Flutter web when using Drift + IndexedDB persistence.

• Platform: Flutter Web (Chrome, multiple tabs)
• Drift version: 2.28.0
• Flutter version: 3.32.7

Problem:
When opening multiple tabs of my app, I sometimes see errors like:

SqliteException(10): while selecting from statement, disk I/O error (code 10)
Causing statement: SELECT * FROM "geolocation_relation" WHERE ...

It seems to happen randomly, even on simple reads, when another tab is also active. My assumption is that this is caused by both tabs trying to access the same IndexedDB-backed database concurrently.

Questions:

1. Is this expected behavior?
2. Is there a recommended strategy for supporting multiple tabs safely?
3. Would retrying failed queries be considered safe, or does it risk data corruption?

Thanks for the awesome work on Drift — would love some guidance on how to handle this scenario properly.

[simolus3]

This is not the expected behavior.

Is this on Chrome Android or another platform? Typically for Chrome we'd use a shared worker to host the database, which should avoid issues with multiple tabs because it's all serialized through Chrome. If you're using drift_flutter could you pass a DriftWebOptions instance with an onResult callback that prints:

1. WasmDatabaseResult.chosenImplementation.
2. WasmDatabaseResult.missingFeatures?

It's easier to give guidance with that information available, since it points at the potential failure.

[Silko5]

Hi, thanks for the guidance.

I added a DriftWebOptions with an onResult callback and got this in Chrome on Windows:
Using WasmStorageImplementation.sharedIndexedDb due to unsupported browser features: {MissingBrowserFeature.dedicatedWorkersInSharedWorkers, MissingBrowserFeature.sharedArrayBuffers}

Based on that, the issue was that Chrome disabled SharedArrayBuffer because my app wasn’t served with the required cross-origin isolation headers.

The fix was to add these headers in my Nginx config:

add_header Cross-Origin-Opener-Policy "same-origin" always;
add_header Cross-Origin-Embedder-Policy "require-corp" always;
add_header Cross-Origin-Resource-Policy "same-origin" always;

After that, Drift started using the shared worker backend as expected, and the random disk I/O errors across tabs went away.

Thanks again for pointing me in the right direction!

[simolus3]

Thanks, but generally sharedIndexedDb should also be a safe storage implementation. I'll take another look to see if this can be improved.