New node entering the cluster on startup invalidates all cluster cache

• Type: Bug
• Status: Closed
• Resolution: No Longer Reproducible
• Affects Version/s: 6.2.X EE
• Fix Version/s: 6.2.X EE
• Component/s: Core Infrastructure, Database Upgrade Framework, Performance > Caching Infrastructure
• Labels:
  • QA-V
  • ee-6.2.x-only
  • ee-ts
• Story Points: 12
• Fix Priority: 3
• Bug Type: Performance

Steps to reproduce

1. Start a node of Liferay Portal 6.2 setup for cluster operation (node1)
2. Monitor it's entity cache for class com.liferay.portal.kernel.dao.orm.EntityCache.com.liferay.portal.model.impl.UserImpl
3. Log in to the first node
   1. Notice that the objects stored are 2 for a clean bundle, but YMMV
4. Start another node of Liferay Portal 6.2 set up for cluster (node2)

Expected result

node1 will maintain its cache.

Actual result

node1 have all its entity cache invalidated as you can see by the number of objects stored going down to 0 again.

Additional details

Using those properties

    # Clear cache on context initialization.
    #
    cache.clear.on.context.initialization=false

    #
    # Clear cache when plugins are undeployed.
    #
    cache.clear.on.plugin.undeploy=false

Will have no effect. Neither will do partially backporting -LPS-58725- to ee-6.2.x.

This is because DBUpgrader.upgrade() is called on startup and it invalidates the cache by calling CacheRegistryUtil.setActive(false).