The following scenarios can be observed due to synchronization failure.
Routes that are deployed on a TalendESB configured in Cellar node do not restart automatically after restarting Cellar node. The feature associated to the Route needs to be manually installed to start the Route. This means that you need to type the feature install command to manually reinstall the route's feature that installs and starts the Route.
When you have the same feature name installed on a cluster group (but with different version), restarting a Cellar node causes overlap between the two features. This causes failure in starting the Service/Route associated to the feature.
There are two different symptoms of the same problem.
An environment consists of three TalendESBs, each on a distinct physical server, and grouped in a single Cellar Cluster.
An environment consists of three TalendESBs (server1, server2, server3), each on a distinct physical server, and grouped in a single Cellar Cluster, Cluster1.
A feature, for example XXXYYService-feature is published and available in the Artifact Repository in two versions, namely 0.1.1 and 0.1.2
Feature version 0.1.2, deployed on a fresh Cluster1, works correctly. The Route/Service associated to this feature is up and running. Restarting one node of the cluster doesn’t cause any issues.
Deploy feature version 0.1.1 onto Cluster1. Restart server1.
The Route/Service associated with the feature version is not more available in server1 or server2.
The problem is linked to the synchronization of the TalendESB when it starts in Cellar Node.
The resolution is to disable the Cellar features sync. This can be done by editing the file in container/etc/org.apache.karaf.cellar.groups.cfg.
Change the following entry:
default.features.sync = false