Allow rolling restart on PortalInstanceLifecycleListener like SamlKeepAliveExpandoPortalInstanceLifecycleListener

• Type: Bug
• Status: Closed
• Resolution: Fixed
• Affects Version/s: 7.0.X, 7.1.X, 7.2.X, 7.3.X, Master
• Fix Version/s: 7.2.X, 7.3.X, 7.3.6 CE GA7, Master
• Component/s: Application Security > SAML
• Labels:
  • ee-ts
• Branch Version/s:
  7.3.x, 7.2.x
• Backported to Branch:
  Committed
• Story Points: 5
• Fix Priority: 3
• Git Pull Request: https://github.com/shuyangzhou/liferay-portal/pull/9630

Rolling restart isn't working when deploying first time listeners like SamlKeepAliveExpandoPortalInstanceLifecycleListener. DXP Cloud rolling restart is an example of so:

Currently rolling restart in DXP Cloud is performed using the following steps:

1. Start a number of new nodes out of a generated image (that can include new applications/modules deployed).
2. Those new nodes will not be visible from the outside, but conform Liferay cluster.

 

Because non Clusterable PortalInstanceLifecycleListener are only executed on master the following can happen:

1. Assume one node is already up.
2. Start a new node from a new image that contains the SAML plugin.
3. PortalInstanceLifecycleListener won't be executed because new node isn't master.
4. Stop the oldest node, previous node becomes master.
5. Start a second node with same image as in step 2.
6. PortalInstanceLifecycleListener won't be executed because new node isn't master.

 

So we end up with a couple of nodes were PortalInstanceLifecycleListener couldn't be executed.