[
https://issues.apache.org/jira/browse/IGNITE-20909?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Pavel Tupitsyn updated IGNITE-20909:
------------------------------------
Description:
Compute jobs can be long-lived and even out-live the client connection. New
Compute API is coming that will return some "execution" object immediately,
which can be used to monitor or cancel the job. Therefore, job startup and
completion should be separated - normal request-response approach is not
suitable.
* Use request-response to initiate the job execution and return an ID to the
client
* Use server -> client notification to signal about completion
This is a tried approach from Ignite 2.x, see linked
[IEP-42|https://cwiki.apache.org/confluence/display/IGNITE/IEP-42+Thin+Client+Compute]
and related discussion
was:
Compute jobs can be long-lived and even out-live the client connection.
Therefore, normal request-response approach does not work well for them.
* Use request-response to initiate the job execution and return an ID to the
client
* Use server -> client notification to signal about completion
This is a tried approach from Ignite 2.x, see linked
[IEP-42|https://cwiki.apache.org/confluence/display/IGNITE/IEP-42+Thin+Client+Compute]
and related discussion
> Thin 3.0: Compute jobs should use server notification to signal completion to
> the client
> ----------------------------------------------------------------------------------------
>
> Key: IGNITE-20909
> URL: https://issues.apache.org/jira/browse/IGNITE-20909
> Project: Ignite
> Issue Type: Improvement
> Components: compute, thin client
> Affects Versions: 3.0.0-beta1
> Reporter: Pavel Tupitsyn
> Assignee: Pavel Tupitsyn
> Priority: Major
> Labels: iep-42, ignite-3
> Fix For: 3.0.0-beta2
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Compute jobs can be long-lived and even out-live the client connection. New
> Compute API is coming that will return some "execution" object immediately,
> which can be used to monitor or cancel the job. Therefore, job startup and
> completion should be separated - normal request-response approach is not
> suitable.
> * Use request-response to initiate the job execution and return an ID to the
> client
> * Use server -> client notification to signal about completion
> This is a tried approach from Ignite 2.x, see linked
> [IEP-42|https://cwiki.apache.org/confluence/display/IGNITE/IEP-42+Thin+Client+Compute]
> and related discussion
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
