A recent post on the AEM forums challenged me to look into an issue I observed myself but did not investigate further.
The observation is that during deployments maintenance tasks are stopped and started a lot, and this triggers a lot of other activities, including a lot of healtcheck executions. This slows down the deployment times and also pollutes the logfiles during deployments.
The problem is that the AEM Maintenance TaskScheduler is supposed to react on changes on some paths in the repository (where the configuration is stored), but unfortunately it also reacts on any change of ResourceProviders (and every Servlet is implemented as a single ResourceProvider). And because this causes a complete reload/restart of the maintenance tasks (and some healthchecks as well), it’s causing quite some delay.
But this behaviour is controlled via OSGI properties, which are missing by default, so we can add them on our own 🙂
Just create a OSGI configuration for
com.adobe.granite.maintenance.impl.TaskScheduler and add a single multi-value property named “
resource.change.types” with the values “ADDED”, “CHANGED”, “REMOVED”.
(please also report this behavior via Daycare and refer to GRANITE-29609, so we hopefully get a fix for it, instead of applying this workaround).