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, though the backend supports partitioning on a column of type int or datetime, we only expose CTID-based partitioning. This is a feature request to support selecting a custom column as the partition key. In the UI, we can show a drop-down for each table, default the value to ctid, and allow the user to select a custom column per table as they wish, only allowing valid types though.
Another important aspect of this is: if a user selects a custom, non-CTID partition key for all the tables as part of this mirror, there is no reason to maintain a long snapshot connection. We create the slot, and then we don't need a long-living connection; we can let the connection close right away since the changes are captured in the slot anyway. This is especially useful in cases where there are many transactions preventing vacuums and causing database issues.
The text was updated successfully, but these errors were encountered:
Currently, though the backend supports partitioning on a column of type int or datetime, we only expose CTID-based partitioning. This is a feature request to support selecting a custom column as the partition key. In the UI, we can show a drop-down for each table, default the value to ctid, and allow the user to select a custom column per table as they wish, only allowing valid types though.
Another important aspect of this is: if a user selects a custom, non-CTID partition key for all the tables as part of this mirror, there is no reason to maintain a long snapshot connection. We create the slot, and then we don't need a long-living connection; we can let the connection close right away since the changes are captured in the slot anyway. This is especially useful in cases where there are many transactions preventing vacuums and causing database issues.
The text was updated successfully, but these errors were encountered: