Closed slafon51 closed 5 months ago
Thanks for reporting this @slafon51 🙏
How many requests per second do you send? Also, is the server responding with 409? (which you could retry on the client)
Is it the first time you encounter this? Or is it recently only? (ie. since we upgraded to SQLAlchemy v2 #3122)
This is something we identified a long time ago https://github.com/Kinto/kinto/issues/602 and at the time we decided not to invest more efforts. This is basically due to the fact that we generate timestamps instead of using sequences
Thank you @leplatrem for your reply.
The server does indeed return a "409 CONFLICT" error. According to the logs I have, we sent an average of 70 requests per second, and got a 409 error every 2 seconds.
Initially, records were created sequentially without any problem. It was only recently that we parallelized it and encountered these errors.
We're going to look at setting up a retry system in the event of 409 errors.
Best regards.
Hello,
During a migration phase, we have a large number of records being created in parallel. These records are created in different collections within the same bucket. We noticed that errors were appearing during the creation of history lines : it seems that the index "idx_objects_parent_id_resource_name_last_modified" generate duplicate keys when 2 history lines are created at the same time.
The error stacktrace :
How can these errors be prevented ?
Best regards.