[jboss-jira] [JBoss JIRA] (WFCORE-3754) Allow ignoring attribute read problems in read-resource calls

[Brian Stansberry (JIRA)]

    [ https://issues.jboss.org/browse/WFCORE-3754?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13563305#comment-13563305 ] 

Brian Stansberry commented on WFCORE-3754:
------------------------------------------

A further thought is a runtime attribute or metric is conceptually different from a configuration attribute. IOW we shouldn't be failing reading config attributes and ignoring that with a warn if it happens is somewhat sweeping something BIG under the rug. As opposed to this whole concept, which is sweeping presumably small things under the rug.

I don't know if that distinction is worth making; just pointing it out. 

> Allow ignoring attribute read problems in read-resource calls
> -------------------------------------------------------------
>
>                 Key: WFCORE-3754
>                 URL: https://issues.jboss.org/browse/WFCORE-3754
>             Project: WildFly Core
>          Issue Type: Feature Request
>          Components: Domain Management
>            Reporter: Brian Stansberry
>
> Occasionally we get reports like JBEAP-14637 where reads of large chunks of data are failing because of a problem with a single attribute.  This RFE is about finding a way to make this more resilient if appropriate.
> I chose Feature Request because I believe this will need to involve user configuration, i.e. we can't just ignore failures by default, with no user control. If we could it would be an Enhancement.
> A PR like https://github.com/wildfly/wildfly/pull/11116 fixes this in a custom way by ignoring the failure (other than logging in server.log) and just issuing a management API WARN. But it would be better to have some sort of control, plus a central implementation.  For example the WARN might be fine for a big recursive read-resource call, but it doesn't make much sense for a read-attribute call.  And it might not make sense for many read-resource calls either. But HAL trying to read the entire /deployment=foo tree might want to configure ignoring failures.
> So I was vaguely thinking about some param to read-resource to control this. And then figuring out how to ignore the failure and WARN in a generic way.



--
This message was sent by Atlassian JIRA
(v7.5.0#75005)