You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
During each transaction commit, we attempt to perform a rollback in case of errors, intending to revert the state from transient to the initial state. However, there is still edge case where the rollback process can fail, specifically when the metadata log index (.query_execution_request_index) is unavailable.
This results in the Flint index becoming stuck in a particular state and subsequent action failure. For example, if a manual refresh fails, and the Flint index is stuck in the refreshing state, any subsequent attempts to manually refresh will also fail until the state is manually updated to active.
What is the expected behavior?
Document the scenario clearly and create SOP that guides user to resolve the issue;
Or consider existing lastUpdatedTime as lease time of the lock (index state)
The text was updated successfully, but these errors were encountered:
What is the bug?
During each transaction commit, we attempt to perform a rollback in case of errors, intending to revert the state from transient to the initial state. However, there is still edge case where the rollback process can fail, specifically when the metadata log index (
.query_execution_request_index
) is unavailable.This results in the Flint index becoming stuck in a particular state and subsequent action failure. For example, if a manual refresh fails, and the Flint index is stuck in the
refreshing
state, any subsequent attempts to manually refresh will also fail until the state is manually updated toactive
.What is the expected behavior?
lastUpdatedTime
as lease time of the lock (index state)The text was updated successfully, but these errors were encountered: