[ 
https://issues.apache.org/jira/browse/FLINK-26281?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17510356#comment-17510356
 ] 

Konstantin Knauf edited comment on FLINK-26281 at 3/22/22, 9:20 AM:
--------------------------------------------------------------------

I've done a couple of tests with ES7, none with ES6: DataStream & Table API, 
NONE/AT_LEAST_ONCE, Elastic Search Index, Elastic Search Data Stream [1]. 

Overall, no blockers. Here are some suggestions for improvements to the 
documentation: 
 * explain how end-to-end exactly-once can be achieved with the ES connector 
(AT_LEAST_ONCE + upserts based on a deterministic id). In my understanding it 
is not recommended anymore in Elastic to set the id manually, so this might be 
worth highlighting. 
 * remove the `type` from the IndexRequest, because it is deprected
 * Add documentation for usage of es datastreams (e.g. requires a @timestamp 
field, only OpType.CREATE, what it means for end-to-end delivery guarantees)



I don't think, I'll have the time to look into ES6. So, [~alexanderpreuss], 
I'll leave it up to you to either move this to DONE or to unassign myself. 

[1] 
[https://www.elastic.co/guide/en/elasticsearch/reference/current/data-streams.html]


was (Author: knaufk):
I've done a couple of tests with ES7, none with ES6: DataStream & Table API, 
NONE/AT_LEAST_ONCE, Elastic Search Index, Elastic Search Data Stream [1]. 

Overall, no blockers. Here are some suggestions for improvements to the 
documentation: 

* explain how end-to-end exactly-once can be achieved with the ES connector 
(AT_LEAST_ONCE + upserts based on a deterministic id). In my understanding it 
is not recommended anymore in Elastic to set the id manually, so this might be 
worth highlighting. 
* remove the `type` from the IndexRequest, because it is deprected
* Add documentation for usage of es datastreams (e.g. requires a @timestamp 
field, only OpType.CREATE, what it means for end-to-end delivery guarantees)

[1] 
https://www.elastic.co/guide/en/elasticsearch/reference/current/data-streams.html

> Test Elasticsearch connector End2End
> ------------------------------------
>
>                 Key: FLINK-26281
>                 URL: https://issues.apache.org/jira/browse/FLINK-26281
>             Project: Flink
>          Issue Type: Improvement
>          Components: Connectors / ElasticSearch
>    Affects Versions: 1.15.0
>            Reporter: Alexander Preuss
>            Assignee: Konstantin Knauf
>            Priority: Blocker
>              Labels: pull-request-available, release-testing
>             Fix For: 1.15.0
>
>
> Feature introduced in https://issues.apache.org/jira/browse/FLINK-24323
> Documentation for [datastream 
> api|https://nightlies.apache.org/flink/flink-docs-master/docs/connectors/datastream/elasticsearch/]
> Documentation for [table 
> api|https://nightlies.apache.org/flink/flink-docs-master/docs/connectors/table/elasticsearch/]
> As 1.15 deprecated the SinkFunction-based Elasticsearch connector and 
> introduces the new connector based on the Sink interface we should test it 
> behaves correctly and as the user expects.
>  
> Some suggestions what to test:
>  * Test delivery guarantees (none, at-least-once) (exactly-once should not 
> run)
>  * Write a simple job that is inserting/upserting data into Elasticsearch
>  * Write a simple job that is inserting/upserting data into Elasticsearch and 
> use a non-default parallelism
>  * Write a simple job in both datastream api and table api
>  * Test restarting jobs and scaling up/down
>  * Test failure of a simple job that is inserting data with exactly-once 
> delivery guarantee by terminating and restarting Elasticsearch
>  * Test against Elasticsearch 6.X and 7.X with the respective connectors



--
This message was sent by Atlassian Jira
(v8.20.1#820001)

Reply via email to