RHQ trait storage

Heiko W.Rupp hrupp at redhat.com
Fri Apr 25 09:50:17 UTC 2014


Am 24.04.2014 um 20:47 schrieb Elias Ross <genman at noderunner.net>:
> -1 on creating a new/separate purge process

We already have a trait purge process 
org.rhq.enterprise.server.scheduler.jobs.DataPurgeJob#purgeMeasurementTraitData
which calls
org.rhq.enterprise.server.measurement.MeasurementDataManagerBean#purgeTraits

That would "just" be rewritten to look in C* instead of the RDBMS.

In theory we could just add a TTL to each trait so that C* just "auto-expires" it.

Question though: what do we do with traits (and this applies to the current expression
as well): what should happen if a trait value just has not changed in a year. If we expire
the existing (only) value, then the server would not show anything. And as the agent is
caching the previous value and not sending it along again (until a restart), there may
be a long time with no value.






More information about the rhq-devel mailing list