[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-ha-dev
Subject:    Re: [Linux-ha-dev] Patch: SNMP-Subagent initial submission
From:       Alan Robertson <alanr () unix ! sh>
Date:       2002-11-14 15:15:02
[Download RAW message or body]

Greg Freemyer wrote:
>  >>  > 
>  >>  > For a simple 2-node cluster with both node A and node B 
>  >>  > operational and node A 
>  >>  > providing the service, is the below desired?
>  >>  > 
>  >>  > snmp poll of CVIP - Resource Operational                   
>  >>  > (resource status on 
>  >>  > primary node)
>  >>  > snmp poll of Node A IP - Resource  Operational           
>  >>  > (Local node resource 
>  >>  > status)
>  >>  > snmp poll of Node B IP - Resource  Down                    
>  >>  > (Local node resource 
>  >>  > status)
> 
>  >>  See, I think here's what I am confused at. I don't know what
>  >>  is the difference between the "resource status on primary
>  >>  node" and "local node resource status on Node A". Shouldn't 
>  >>  they be the same since Node A is providing the service? 
> 
> You are correct that they give the same status in the above, but now assume I 
> move the CVIP and associated resources to Node B.
> 
> Now you have:
> 
> CVIP - up
> node A - down
> node B - up
> 
> So, you interrogate the CVIP when you want to know if the resource is available 
> at all.  You interrogate the nodes IP when you want to know where it is 
> running.
> 
> Another more complex example is DRDB.  This has a master process on the primary 
> server and a replicator (or standby) process on the non-primary server.
> 
> So now you have 2 "normal" situations:
> 
> CVIP - up
> node A - up
> node B - standby
> 
> CVIP - up
> node A - standby
> node B - up
> 
> and 2 degraded, but fully functional situations:
> 
> CVIP - up
> node A - up
> node B - down
> 
> CVIP - up
> node A - down
> node B - up
> 
> In the above, it is only by interrogating the CVIP, that you always get the 
> "right" answer from a cluster-wide perspective.  To get degraded status info 
> however, you have to interrogate the 2 nodes independently.
> 
> As you can see, you really do want to have the snmp agent available on the node 
> IPs and on the CVIP.
> 
> What I don't know is how you handle snmp traps.  I would think that if the node 
> is primary then you send the trap from the CVIP, and if the node is non-primary 
> you send it from the node IP.

By the way, I keep hearing the term "primary".  Heartbeat has no concept of 
"primary" at all - particularly not a primary node.  The term primary might 
apply to certain active/passive configurations, but heartbeat isn't aware if 
you configure your system that way.

A *service* may be active on at most one machine.  I suppose you could call 
that machine "primary" for that service if you wanted to.  But, it's not 
"the primary node" or anything like that.

	-- Alan Robertson
	   alanr@unix.sh

	-- Alan Robertson
	   alanr@unix.sh

_______________________________________________________
Linux-HA-Dev: Linux-HA-Dev@lists.community.tummy.com
http://lists.community.tummy.com/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/
[prev in list] [next in list] [prev in thread] [next in thread]