On 9 May 2015, at 18:33, Gary Brown wrote:
I think we probably need to revisit the use cases for the event store
Yes, please start with this. A good number of the previous messages
dealt with
implementation details of a sub use case.
We may have several types of the (abstract) "Event" like
- Audit record
- Fired Alert
- BTM call results
- remote syslog
- ...
that all have their specific details. With a header field in the Bus
messages
we can differentiate them later for further processing.
We need some degree of separation here, as e.g. Audit records are
append-only.
The user should not be able via normal means to delete (individual)
records, while
for other types, we may care less.
Also for Audit records I see a need to persist them to some external
storage before
expiring them, which may not be true for other types.
Within the Hawkular family of projects, I also think, this should be a
new separate
(logical) component, that may even use the common C* store as Inventory,
Metrics and
Alerts already do.
We should nevertheless not merge it into the others, as the requirements
are much
different.