[ 
https://issues.apache.org/jira/browse/FLINK-41212?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Timo Theusner updated FLINK-41212:
----------------------------------
    Description: Without a workflow scheduler, create, convert, suspend, resume 
and drop of a FULL refresh mode materialized table fail with "The workflow 
scheduler must be configured …". The message doesn't say that continuous mode 
still works or how to get it. Table API programs hit this when 
{{materializeInto}} gets a freshness above the threshold, because the AUTOMATIC 
refresh mode then resolves to FULL. Extend the messages so they state the 
limitation and the remedy.  (was: {{FULL}} tables are refreshed by a 
{{{}WorkflowScheduler{}}}, which only the SQL Gateway hosts, so a plain 
{{TableEnvironment}} cannot schedule them. Reject creating or converting a 
table that resolves to {{FULL}} there before any catalog write, and reject 
{{{}SUSPEND{}}}, {{RESUME}} and {{DROP}} of existing {{FULL}} tables with an 
error naming the SQL Gateway.

Proposed Changes
 * Reject in {{DefaultMaterializedTableExecutor}} a create or conversion that 
resolves to {{FULL}} without a workflow scheduler, declared or via 
{{{}materialized-table.refresh-mode.freshness-threshold{}}}, before any catalog 
write
 * Reject {{{}SUSPEND{}}}, {{RESUME}} and {{DROP}} of a {{FULL}} table without 
a workflow scheduler, naming the SQL Gateway in the error
 * Allow {{ALTER}} (catalog only) and {{REFRESH}} (one-off batch job) of a 
{{FULL}} table without a workflow scheduler)

> Explain FULL refresh mode rejections without a workflow scheduler
> -----------------------------------------------------------------
>
>                 Key: FLINK-41212
>                 URL: https://issues.apache.org/jira/browse/FLINK-41212
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Table SQL / Planner
>            Reporter: Timo Theusner
>            Assignee: Timo Theusner
>            Priority: Major
>              Labels: pull-request-available
>
> Without a workflow scheduler, create, convert, suspend, resume and drop of a 
> FULL refresh mode materialized table fail with "The workflow scheduler must 
> be configured …". The message doesn't say that continuous mode still works or 
> how to get it. Table API programs hit this when {{materializeInto}} gets a 
> freshness above the threshold, because the AUTOMATIC refresh mode then 
> resolves to FULL. Extend the messages so they state the limitation and the 
> remedy.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to