[
https://issues.apache.org/jira/browse/FLINK-38032?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18005068#comment-18005068
]
Qilong Wang commented on FLINK-38032:
-------------------------------------
I have conducted cross testing on FLIP-519, and the verification steps are as
follows:
# Checkout and compile flink in commit branch: release-2.1.
# Start a standlone flink cluster.
# Create sql-script in the sql-client.
# I have prepared two data input tables and one data output table. An input
table is using a fileSystem connector that supports LookupTableSource.
# Use 'SET 'table.exec.async-lookup.key-ordered-enabled' = 'true';' and 'SET
'table.exec.async-lookup.output-mode' = 'ALLOW_UNORDERED';' to set necessary
conf.
# Add a agg operator ahead of the lookup join to generate cdc changelog.
Simultaneously using lookupJoin to connect another table.
# For Test 1, use explain to check that the LookupJoin operator does contain
information for asyncs=[ORDERED, KEY_ORDERED: true, 180000ms, 100].
# Regarding Test 2, check the results. It was indeed found in the TM of Flink
UI that the results under the same upsert key is ordered.
> Release Testing: Verify FLIP-519: Introduce async lookup key ordered mode
> -------------------------------------------------------------------------
>
> Key: FLINK-38032
> URL: https://issues.apache.org/jira/browse/FLINK-38032
> Project: Flink
> Issue Type: Sub-task
> Reporter: Shuai Xu
> Priority: Major
>
> Follow up the test for https://issues.apache.org/jira/browse/FLINK-37874
> We now support a async key ordered lookup join that performs a lookup join in
> the order of specified key.
> Now we'd like to have some cross team testing to validate it. We want to
> cover the following cases
> Test 1: Enable async key ordered lookup join, confirm the lookup join
> operator in flink UI contains contents like 'async=[UNORDERED, KEY_ORDERED:
> true, 180000ms, 100])'. 'KEY_ORDERED: true' means the optimization is enabled.
> Test 2: Verify the correctness of result. This optimization requires
> non-append-only stream. For simplity, add a agg operator ahead of the lookup
> join to generate cdc changelog and the upsert key which would be refered as
> the order key in async key ordered lookup join. You could use the filesystem
> connector for source table and lookup table.
> Here are details for Test 2:
> Step1. setup the envirionment and prepare the source data and lookup data.
> Step2. create sql-script in the sql-client.
> Step3. use [explain
> statement|https://nightlies.apache.org/flink/flink-docs-master/zh/docs/dev/table/sql/explain/]
> to find the upsert key before lookup join operator.
> Step4. check the result. Ensure the result under the same upsert key is
> ordered.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
