We discovered a few interesting things about consolidation tasks and HFM 18.104.22.168 at a recent client engagement. First, there is no “keep-alive” between HFM and the Oracle HTTP Server, so anything relying on a keep-alive will time out and cause issues later. Load balancers would need to have their keep-alive times adjusted, and OHS would need the entry for WLIOTimeoutSecs in the mod_wl_ohs.conf file adjusted upwards too.We saw this issue with our install due to Task Flows that incorporated consolidation tasks. If a task in our environment took more than 5 minutes to complete, a new identical task would be spawned. The consolidation CPU requirements were exceeding the capabilities of the HFM servers because of this. We adjusted the LB timeout and got our test task to not duplicate. However, it only took 23 minutes to complete. We found the OHS timeout (7200 seconds, or 2 hours) in our environment when we ran longer consolidation tasks. We upped the timeout to 16 hours in OHS and have not seen a repeat of the issue.
In addition to the extra task being spawned, the task takes up a connection in the EPMSystemRegistry pool in WebLogic - default is 30, so if you have a LOT of consolidation tasks, you can exceed this value and cause failure for other tasks that require a connection to that pool.
The plan is for Oracle to introduce a keep-alive in HFM in version 22.214.171.124.300, scheduled for release in November 2017.