[
https://issues.jboss.org/browse/TEIID-4626?page=com.atlassian.jira.plugin...
]
Steven Hawkins commented on TEIID-4626:
---------------------------------------
Here's some possible refinements:
https://github.com/shawkins/teiid/commit/6f8b67d097f9293802c9548bc0c117ec...
- discover a unique set of contributing tables
- remove the need for refresh type property
- allow for both ttl and staleness to be used at the same time
- refined the staleness calculation to account for cardinality 0 and to be based upon a
percentage of the square root rather than just the percentage. So a likely recommendation
would be 10 which would mean 1 update for a 100 row matview, 100 updates for a million row
matview, 1000 updates for a 100 million row matview, etc.
Add a full snapshot refresh strategy based upon table modifications
-------------------------------------------------------------------
Key: TEIID-4626
URL:
https://issues.jboss.org/browse/TEIID-4626
Project: Teiid
Issue Type: Enhancement
Components: Query Engine
Reporter: Steven Hawkins
Assignee: Ramesh Reddy
Fix For: 9.3
using the existing full materialization refresh logic we can trigger a refresh based upon
a notion of how dirty a table is based upon a proportion of rows updated to the rows in
the table. This would also require some change to the status table to capture the number
of transitive row updates. This could then be hooked up to debezium as the event source.
--
This message was sent by Atlassian JIRA
(v7.2.3#72005)