[
https://issues.apache.org/jira/browse/IGNITE-4699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15953085#comment-15953085
]
Vladimir Ozerov commented on IGNITE-4699:
-----------------------------------------
[~sharpler],
There are two main problems which I expect to be covered once ticket is ready:
1) Execution of nested compute calls. In this case you can make decision on
per-job basis. Say, if you created a re-usable component, you could do
something like:
{code}
compute.withExecutor("component_executor").[run_job]
{code}
And any code on top of this component will be able to use it safely without a
risk of deadlock.
2) Execution of chained calls, especially {{IgniteFuture.chain}} and
{{IgniteFuture.listen}}. Currently user has no control on where continuation is
to be executed, so he can inadvertently run something in public pool, causing
exhaustion of public pool threads. Once user has access to predefined
executors, he will be able to control this behavior.
Regarding simplicity of configuration, for other thread pools we allow users to
configure only number of threads (e.g.
{{IgniteConfiguration.publicThreadPoolSize}}. And I am not aware of any
complains on insufficient flexibility of this approach. This is why I propose
to start with as simple config as possible. We will always be able to extend it
in future if needed.
One more important reason why we need to have this feature ready ASAP is 2.0
release, which is expected to happen in April. This feature will require
message protocol changes, and hence broke binary compatibility.
Having said that I'd prefer to go ahead with Taras patch at the moment, while
keeping in mind all proposed improvement for future releases. Ideally, it would
be cool to materialize them into JIRA tickets, and plan their implementation in
AI 2.1 release or so.
> Introduce custom configurable executors.
> ----------------------------------------
>
> Key: IGNITE-4699
> URL: https://issues.apache.org/jira/browse/IGNITE-4699
> Project: Ignite
> Issue Type: Task
> Components: compute
> Reporter: Vladimir Ozerov
> Assignee: Alexander Menshikov
> Labels: important
> Fix For: 2.0
>
>
> We need to provide a way to configure custom thread pools for user compute
> tasks.
> Proposed API:
> 1) Config
> {code}
> class ExecutorConfiguration {
> String name;
> int size;
> }
> {code}
> 2) Choosing executor for compute task:
> {code}
> IgniteCompute compute = Ignite.compute().withExecutor("my_exec");
> {code}
> 3) Accessing raw executor (could be required for proper execution of future
> listeners):
> {code}
> ThreadPoolExecutor exec = ignite.compute().localExecutor("my_exec");
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
