Threading/Deadlock/Concurrency Issue? about sqlstreamstore HOT 8 CLOSED

I've worked on this with @wooderz privately. We found out that his Guid based stream identity generation logic was at fault. Every so often it would generate a duplicate value (where a unique one was expected) and cause the expected behaviour around appending to a stream with ExpectedVersion.NoStream to kick in.

from sqlstreamstore.

It's a known issue with the MySql implementation, but it's news to me with the Postgres one ... I'll try and add a test to reproduce the behavior. As of how much concurrency are you observing this?

from sqlstreamstore.

Wasn't lots really, about 120 I think. I use Loadster to do the tests, here's the page with the error on. Each test is set to create a new record (using a random identifier in the post body) so each one should be a NewStream

from sqlstreamstore.

To be 100% sure I just ran it with a dataset of all unique values for keys for the tenant id and the same problem occurred. Therefore it's definitely not an ID conflict

from sqlstreamstore.

Great, was going to ask - makes looking for the needle in the haystack easier since there should be a needle.

from sqlstreamstore.

Anything I can do to narrow it down? Happy to share my aggregate Save method if that's of any help

from sqlstreamstore.

If you can share it, that might shed some extra light.

from sqlstreamstore.

(FWIW Equinox.Tool (see https://github.com/jet/equinox#benchmarks and the section just above) is plumbed to run a concurrency test with error and latency metrics using SqlStreamStore - I have not personally used it for some time and don't have much direct interest in mucking with SQL stores atm)

from sqlstreamstore.