[
https://issues.apache.org/jira/browse/SLING-9211?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17160917#comment-17160917
]
Georg Henzler commented on SLING-9211:
--------------------------------------
I analysed the problem and I think the problem only exists on first startup
(subsequent startups seem fine). The following is happening:
* the OSGi framework quickly "shoots up" to start level 30
* then the OSGi installer installs the bundles following the "conceptual start
level order" as given by provisioning files but at actual start level 30 (also
the start level is set to bundle for subsequent startups, but first startup
installs everything on start level 30)
* only once all bundles are installed the configurations are installed (there
is a significant delay between installing a bundle and configurations that are
meant for the bundle).
* While the configurations for ServiceUnvailableFilter and systemready health
checks are not available yet (delay from point above), the Sling startup page
is not shown yet
There are multiple options to fix the problem:
# Extend health check core with a bundle activator that "forces the relevant
configurations" into the systems
** I have tested this approach (using "felix.health.ServiceUnavailableFilter.*"
framework properties) and it works, however it feels like a workaround
# Change the launchpad to generally install configurations straight after the
config admin is available (this seems conceptually the right thing to do for
me, but there might be reasons for the current state that I'm not aware of)
# Change the launchpad to mark certain configurations as "bootstrap" (or
similar naming), only those are installed immediately after the config admin
becomes available
I also see that SLING-9498 introduced the possibility to run it with the
feature model (maybe there configurations are respected as early as possible
during startup? I haven't tried yet)
[~cziegeler] WDYT, what option from above would be best?
> Startup detection not working reliable anymore
> ----------------------------------------------
>
> Key: SLING-9211
> URL: https://issues.apache.org/jira/browse/SLING-9211
> Project: Sling
> Issue Type: Improvement
> Components: Starter
> Reporter: Georg Henzler
> Assignee: Georg Henzler
> Priority: Major
>
> It seems with SLING-8418 the 503 responses are not as reliable as with the
> previous custom solution.
> To be checked with script in
> https://issues.apache.org/jira/browse/FELIX-6097?focusedCommentId=16813130&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16813130
--
This message was sent by Atlassian Jira
(v8.3.4#803005)