Unicenter reliability...

[NCT]

Hi,
I'm (very) new to Unicenter and have a 'high-level' question.
I'm a bit worried about the possibility of loosing fault messages.
CA say they offer guaranteed delivery to a proprietary binary log file on a DSM, but after that I don't see how I can ensure the message ends up in a database of some sort.

Is it possible to have something like guaranteed delivery of a detected fault into a central datawarehouse (Tivoli say the have guaranteed delivery from agent to a DB2 datawarehouse).
Basically, I need to be sure a fault message cannot be lost.

Or, does anyone have any idea on information which might describe the details of how a message flows from an agent to wherever it goes?

Any help appreciated.

 

[j1bber]

Depending upon the "message" this would normally indicate a change of state of the agent. This change of state is normally replicated to the CORE (SQL database) that contains the current state of all managed objects via the worldview gateway on the DSM.
If this isn't certain enough for you you could allways monitor specific agents from multiple dsm's
Agents can also be configured to send SNMP traps to multiple destinations.
Remember whatever you use SNMP is not a reliable or secure protocol.

 

[JonnoGT]

This is going to be a long one - sorry

There are 3 main components:
Agent - DSM (agent manager) - Core

When an event occurs on the agent it sends the event via SNMP to the DSM, the DSM then forwards it to the Core console. CA uses it's own version of SNMP on a dedicated port - so although there is no guarantee of delivery I have never yet lost a message (I think!).
The DSM has a facility called store and forward which guarantees that messages will be delivered to the core. If for example the core is down, the DSM will queue messages and send them when the core becomes available again.

Also, the CA event and workload agents send messages directly to the CORe and also includes the Store and Forward (SAF) functionality. The CA event agent basically captures all events that happen on a server. EG all NT event log messages, all syslogd messages etc, and forwards these to the CA console on the CORe.

One important point to make is that events in CA do not get sent to a database. They go to the CA console which is a flat file. The objects status is updated in the core DB, but the actual event message goes no-where near a DB.

The flow is as follows:
SNMP Agent -to- SNMP gateway -to- DSM - Interprets trap, Puts message on Console (flat log file can be on DSM or Core server. Uses SAF if needed) -to- World View Gateway - updates agent status in Core DB.

Event Agent -to- Console - using CCI (guaranteed delivery due to SAF functionality).

Because the standard CA agents use SNMP, if the DSM goes down there is no gaurantee of delivery. When CA say "they offer guaranteed delivery to a proprietary binary log file on a DSM" they mean between the DSM and the Console - not between an agent and the DSM.
--------------------------------
An important point is: the CA console is inclusive by default. IE all messages go to it.
The Tivoli console (and others) are exclusive by default. IE you have to send selective messages to it.

You would need a big DB to be able to store all the events in an inclusive console. It's not really possible.

The CA console is by far and away the best on the market. Other products have better or more configurable agents, but the CA console is miles ahead in terms of usefulness, useability and pretty much everything. For finding the source of and diagnosing a problem nothing beats it. Because it is inclusive you can correlate events and see what the root cause of a problem was. The fact that it doesn't go to a DB has never been an issue for me.

I also doubt any vendor that claims to be able to guarantee end to end delivery. CA, IBM or any of them. If your server goes down, you'll always loose data.