[
https://issues.apache.org/jira/browse/SLING-5888?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15424126#comment-15424126
]
Bertrand Delacretaz commented on SLING-5888:
--------------------------------------------
Now that I understand the principle better, I think the messages sent to the HC
log should more directly expose the installer state, so that someone
troubleshooting their instance becomes more familiar with how the installer
works and to help make the whole thing more transparent.
For example instead of saying "The OSGi bundle for '{}' was not installed" I'd
prefer "The installer state of the OSGi bundle resource {} is {}, probably
because a later version of that bundle is already installed" or something like
that - exposing the {{ResourceState.IGNORED}} and {{ResourceState.INSTALL}}
values, to help users relate to other webconsole pages which show those values.
And for configs something like "The installer state of the OSGi config resource
{} is {}, config might have been manually overwritten".
And maybe refer to "the {{org.apache.sling.installer.healthcheck}} 's Sling
documentation page" for details and hints? It's easier to enhance that page
rather than try to include too much info in the log messages. That might be an
additional HC log message emitted only if there are other "red flag" messages.
(BTW the use of both {{log.}} and {{LOG.}} variables is not my favorite ;-)
maybe use hcLog. for the HC log)
Does that make sense?
Sorry to be picky on this but I think the general problem is far from obvious
for someone who's not familiar with the installer internals, so providing
precise information and good documentation will help.
> Health Check for detecting not-installed configurations/bundles from the OSGi
> installer
> ---------------------------------------------------------------------------------------
>
> Key: SLING-5888
> URL: https://issues.apache.org/jira/browse/SLING-5888
> Project: Sling
> Issue Type: Improvement
> Components: Health Check, Installer
> Reporter: Konrad Windszus
> Assignee: Konrad Windszus
> Attachments: SLING-5888-v01.patch
>
>
> We faced an issue in the past that configurations (or bundles) being deployed
> in the repository are not effective (i.e. not installed in Apache Felix).
> The reason for that is that the OSGi configuration was either manually
> modified in the Web Console (and therefore takes precedence over the version
> of the configuration being deployed in the repository) or in case of bundles,
> that the same bundle has already been deployed in that or a newer version
> (manually or through some other path in the repository/filesystem).
> Both states may lead to issues at run time, so it would be good to have an
> automated check for it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
