SNMP traps from multihomed host

[chojin]

I'm trying to retrieve SNMP traps for clariion boxes. Those boxes are each defined as one host, using 2 IP addresses (one for each storage processor).
Each of those storage processors send SNMP traps using their own IP.

However despite the fact that the host in zabbix is configured with 2 snmp interfaces, only snmp traps for the default snmp interface are recognized by zabbix. So we only receive traps from one storage processor.

I understand that each host-item is connected to one specific host-interface. So the items catching those traps, are associated with the default host snmp-interface (the first one)

Hence I try to Clone those items to assign them to the other host-interface to be able to also catch the traps sent by the other storage processor. But it seems that I can't since the item-key would then be duplicate.

Am I missing something or is it plain not possible to receive traps from two IP adresses from the same host?

[chojin]

*bump* anyone?

[chojin]

*bump* - Still struggling with this problem.. Does anyone have an idea?
I know that creating seperate hosts for Service Processor (SP) A and SP B would solve this for the snmp traps, however other metrics would then need to be defined on only one of those hosts or I should create a thrid host for those which would clutter the zabbix hosts list..

[Smersh]

Is there the common SNMP interface on clariion box? If yes, then in theory, you can use snmp discovery to retrieve all its ip addresses (I'm using perl script to similar task for cisco routers, as to each interface can be assigned several IPs).

[chojin]

There is no such thing as a common (SNMP) interface. There are 2 'storage processors' load-balancing the storage traffic. Both have an interface on which they answer SNMP requests. System-wide metrics can be queried on both SP's and they will both return the same value's. They are both an interface to the same storage box and one can fail, the other will seamlessly take over all activity.

Both can also send SNMP traps depending on which SP the event occurred. Sometimes a system-wide event is sent as a snmp trap by both processors simultaneously.
And my problem is those snmp traps: I can only define one unique snmptrap-item within a host and this snmptrap item is bound to interface one OR interface two of the host. So I can only capture snmp-traps from one storage processor. And I can't clone the snmptrap-item to listen to the other host-interface as then the item-key is found to be duplicate within the host.