Im currently testing some code to know which data comes from failing node and I wonder if there is some pattern that I missed on google (cant search jb forums due to: the module threw an exception ...)
To make it short, scenario looks as follows:
- threee nodes, sharing state via jb cache
- service unit running on node1 creates timers
- node1 fails
- some service is notified of failure (either via HA Partition or cache listener) and node2 is peaked up as one that takes over
- now node2 should instantiate timers from failed node, however how can it distinct one timer(cache data describing it) from another (from its local, or node3 timers)? - unless there is another cache structure indexing data by node, which seems a bit bad.
Currently Im testing cache listeners to see if it can give some valued information.
Any one had similar case?