[PHOENIX-7381] Client should not validate LAST_DDL_TIMESTAMP for a table with non-zero UPDATE_CACHE_FREQUENCY if the client's cache entry is not old enough

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 5.3.0
• Assignee: Palash Chauhan
• Reporter: Palash Chauhan
• Created: 12/Aug/24 21:13
• Updated: 01/Oct/24 06:37
• Resolved: 01/Oct/24 06:37

Description

PHOENIX-6883 introduced a way for a client to validate its table metadata before queries or updates using LAST_DDL_TIMESTAMP of the table. In the current implementation, client will always perform an RPC to a random region server to validate timestamps since we would set DEFAULT_UPDATE_CACHE_FREQUENCY to NEVER. 

However, if a client operation involves a table with a positive UPDATE_CACHE_FREQUENCY defined in its schema, we should skip the timestamp validation if the client's cache entry for the table is not old enough. (just like we skip getTable rpc if the cache entry is not old enough)

Issue Links

• links to: GitHub Pull Request #1952