Bug 1357003

Description of problem:
When connecting Cloud Forms 5.6.0.13 to a RHEVM environment that has data warehousing installed and configured (ovirt_engine_history) there is a spike in memory usage resulting in reset workers and much higher than "normal" resource usage during the initial C&U collections.

The cause of this issue stems from that fact that the ovirt_engine_history database can contain > 24 hours of realtime sample in which the collector workers will collect all potential realtime data samples they can obtain.  This not only spikes the collector workers memory but also results in a spike of perf_rollup messages causing processor workers to stay on the cpu for longer than expected than if connected to another infra provider type of similar size with C&U turned on. (In comparison to VMware environments)


Version-Release number of selected component (if applicable):
5.6.0.13

How reproducible:
Should be almost always reproducible unless your ovirt_engine_history database is brand new and hasn't accumulated many realtime samples.

Steps to Reproduce:
1. Connect CFME to RHEVM environment
2. Turn on C&U
3. Observe System performance CPU & Memory of the CFME appliance

Actual results:
Depending on environment size and when the last time a "purge" job has run on the ovirt_engine_history database will vary the amount of memory growth and resulting number of reset workers in CFME

Expected results:
Collector workers to collect C&U data without huge spikes in memory growth during the initial collection/processing.

Additional info:
Previous versions of this same issue are documented here: https://bugzilla.redhat.com/show_bug.cgi?id=1085988

Potentially the capture of realtime metrics from beyond an hour can be viewed as a feature of CFME managing RHEVM as we capture more historical samples compared to VMware which only stores 1 hour of realtime samples.