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
Currently, on performance grounds, xline provides a fast-read mechanism(ReadState) to a client, which can prevent a leader node from becoming a bottleneck in an xline cluster. Thanks to the implementation of the read-state part, a command has to go through three stages from prepare to execute and then to after_sync, just for some notification reasons, even though it fails in its earlier stage. This doesn't make sense. We should refactor the implementation of the ReadState part.
The text was updated successfully, but these errors were encountered:
Passing LogIndex to the prepare and execute stage of a command can trigger barriers in advance when failures occur, and it is no longer necessary to call other stages for failed commands.
Currently, on performance grounds, xline provides a fast-read mechanism(ReadState) to a client, which can prevent a leader node from becoming a bottleneck in an xline cluster. Thanks to the implementation of the read-state part, a command has to go through three stages from prepare to execute and then to after_sync, just for some notification reasons, even though it fails in its earlier stage. This doesn't make sense. We should refactor the implementation of the ReadState part.
The text was updated successfully, but these errors were encountered: