[ 
https://issues.apache.org/jira/browse/IGNITE-9607?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16728749#comment-16728749
 ] 

Vyacheslav Daradur commented on IGNITE-9607:
--------------------------------------------

[~agoncharuk], thank you for the notes!
{quote}1) Not sure if it was discussed with other community members, but I 
think it may be better to get rid of various {{instanceof}} statements in the 
code and have a common interface for both service processor implementations 
(possibly, noop)
{quote}
Previously such approach has been used, but it has been reworked during a 
review. In the main code base (not in tests) we use 'instance of' in rare 
operations, mostly at startup of kernal or its components and on local join 
event, such checks are not called in heavy loaded places.
Also, we can't get-rid 'instance of' completely, e.g. 'onActivate' and 
'onDeactive' should be called in different places depending on service 
processor implementation. I'd like to leave it as is if you don't mind.
{quote}2) Should we narrow down the generic type of ServiceDeploymentFuture to 
Serializable?
{quote}
Makes perfect sense to me. Done.
{quote}3) In IgniteServiceProcessor#stopProcessor() we need to wrap 
fut.onDone(stopError) in a try-catch block. We recently discovered that 
onDone() call can re-throw the exception to the caller, which prevents correct 
node stop
{quote}
Done.
{quote}4) I think we should add (maybe in a separate ticket) some diagnostic 
mechanics to dump pending deployment futures. The best option is to use 
existing PME diagnostic mechanics
{quote}
I've filled a task IGNITE-10817

 

> Service Grid redesign - phase 1
> -------------------------------
>
>                 Key: IGNITE-9607
>                 URL: https://issues.apache.org/jira/browse/IGNITE-9607
>             Project: Ignite
>          Issue Type: Improvement
>          Components: managed services
>            Reporter: Vyacheslav Daradur
>            Assignee: Vyacheslav Daradur
>            Priority: Major
>             Fix For: 2.8
>
>
> This is an umbrella ticket for tasks which should be implemented atomically 
> in phase #1 of Service Grid redesign.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to