When a new benchmark is imported, metrics are recalculated, which can be a lengthy process for large deployments.
No feedback is provided until recalculation is complete. This came to our attention because in a deployment w/ nginx that cut off the connection because no response was provided within the default proxy_read_timeout default of 60s.
Consider:
Streaming a response showing progress being made to prevent response timeout issue
When a new benchmark is imported, metrics are recalculated, which can be a lengthy process for large deployments. No feedback is provided until recalculation is complete. This came to our attention because in a deployment w/ nginx that cut off the connection because no response was provided within the default proxy_read_timeout default of 60s.
Consider: Streaming a response showing progress being made to prevent response timeout issue
User also reports getting an "Error: The number of locks exceeds the lock table size." error. Research indicates increasing
innodb_buffer_pool_size
and/or locking the entire table when processing a large transaction, rather than letting innodb's row locking handle it: https://stackoverflow.com/questions/6901108/the-total-number-of-locks-exceeds-the-lock-table-size