RHQ Agent's communicate only with RHQ Servers.  The data the agents collect is reported to a server, potentially processed in various ways, like for alerting, and then stored.  Today, availability data, events and more are stored into the [oracle] RDB.  Metric data is stored into the storage cluster you've created.   Agents never communicate directly to the storage cluster.  The storage cluster knows only about storing and fetching data.  It has no concept of agents and all the associated things that go with that, like authorization, communication, etc.

So, RHQ Servers are one thing and they connect to the RDB and the RHQ Storage Node cluster.  They also service agents.  Agents only communicate with an RHQ Server, and not necessarily the server that initially contacted, but rather the server they are told to contact in order to distribute agent load.

I'm sorry, I don't know about F5, but basically the agents are not dealing with F5 after perhaps the initial server contact.  They are connecting to a server via the RHQ comm layer, directly.


On 8/11/2014 2:17 PM, barry.barnett@wellsfargo.com wrote:

Why wouldn’t the agent that points to the F5 GTM URL point to wherever the F5 wants to have it communicate with?  So you’re saying the BCP RHQ server storage node has to join the Prod storage node cluster for this failover to BCP to work, even with the F5 GTM URL being used?

 

Regards,

 

Barry

 

From: rhq-users-bounces@lists.fedorahosted.org [mailto:rhq-users-bounces@lists.fedorahosted.org] On Behalf Of Jay Shaughnessy
Sent: Monday, August 11, 2014 2:07 PM
To: rhq-users@lists.fedorahosted.org
Subject: Re: F5 GTM - RHQ Agent configuration issue

 


If I understand this correctly, the reason is this.  The agent initially contacts *any server* but is not immediately serviced by that server.   It is only provided with a "failover list" of servers it should use.  It then tries to connect with the first server in that list (the "primary").  Failing contact it will then proceed to "failover" to the next HA server in  the ordered list of HA servers.   Eventually it will always try to return to its primary server, if possible.

The failover list will contain entries for only the registered HA servers, in the case A and B. 

On 8/11/2014 1:40 PM, barry.barnett@wellsfargo.com wrote:

We currently have the following configuration for our RHQ server:

 

Production RHQ Server 1 (Node A)

Production RHQ Server 2 (Node B)

·         Both nodes connected to the Oracle backend database)

·         Storage nodes on each instance, clustered

 

BCP RHQ Server (Node C – different location)

·         Storage node not in cluster

 

We have an F5 URL that sits infront of all the nodes, and can point to any of them in the following scenario:

·         Prod Node A fails, traffic directed to Prod Node B

·         Prod Node A and B fails, traffic directed to BCP Node C

 

When an RHQ agent on a client windows box is configured, and the F5 GTM URL is used as the RHQ Server, the agent connects to Prod Node A (as that node is the primary active node for the F5 to point to).  If I bring down both Prod Nodes, I would think the F5 URL would connect the agent to the BCP Node C RHQ server.  But, it never connects to BCP, but rather in its logs shows it trying to connect to both Prod nodes and failing.  Is this due to the storage nodes in Prod being clustered (assumption)?  Why wouldn’t the F5 URL direct the agent to connect to BCP?

 

 




_______________________________________________
rhq-users mailing list
rhq-users@lists.fedorahosted.org
https://lists.fedorahosted.org/mailman/listinfo/rhq-users

 



_______________________________________________
rhq-users mailing list
rhq-users@lists.fedorahosted.org
https://lists.fedorahosted.org/mailman/listinfo/rhq-users