Modify retention policy per topology type to keep 4 hours of raw observation data for finer granularity (4244693)

Modify retention policy per topology type to keep 4 hours of raw observation data for finer granularity

Retention policy for the observation of the topology type has been set to age the initial roll-up period after 4 hours but it is still getting aggregated after 15 minutes. How to set the retention policy per topology type to keep 4 hours of raw observation data for finer granularity?

The issue here is probably that there are too many values for the observation, so Foglight Management Server (FMS) is aggregating them earlier. The metrics seem to be sampled every minute and the retention policy is configured to keep values in memory for 4 hours before persisting them. That means that there would be about 240 values in memory at any time. The cache service is currently configured to only allow 30 transient values for any observation, and will aggregate the values if there are more values than that. This would be why the 15 minutes aggregation is performed earlier than expected.

Having an age of 4 hours for the age of the initial rollup period is not recommended. The age should really be an hour or less to ensure that values are persisted without too much delay and can be processed by the nightly roll-up, etc.

How to set the retention policy to keep 4 hours of raw observation data for finer granularity?
The retention policy period that says "age 4hours; rollup period 15min" could be changed to "age 1min; rollup period 1min" so that the server persists the 1 min data. The "age" can be quite small (i.e. 1 min) because little or no aggregation is being applied to the values, so there will not be any need to preserve the raw values for a longer period of time. The 1 min aggregate values would then be available for 3 days. This retention policy should be fine so long as it is applied selectively. If the server starts persisting 1min aggregates for a very large number of observations then that would increase the load on the database.