We are facing a weird situation with our physical HP servers in NNMi, and I would like to know if someone has the same problem, and if someone managed to fix the problem.
Briefly, NNMi merges HP servers and ILO mgmt cards into one node, and the server's IP address seems to be removed from NNMi.
We tried the following:
1- Delete all nodes having this problem from NNMi
2- Add all IP addresses assigned to ILO mgmt card to the Exclude IP Addresses list (Discovery Configuration) page
3- Seed the servers using their IP address (or using their FQDN)
After doing that, we were able to see the servers, the IP, and the connection to the switch correctly (SNMP polling works fine, ICMP works fine). The servers look OK for a small period of time, and then NNMi puts back the ILO IP address, it changes the display name (FQDN of the ILO card), and the IP address assigned to the server (the one we manually seeded) disappear from NNMi.
For example, we have a physical
Windows 2k8 server:
sli1793.sli.bz => with IP address: 10.160.203.229
ILO Mgmt card:
ribsli1793.sli.bz => with IP address: 10.160.203.230
The initial discovery of NNMi, discovered ribsli1793.sli.bz (managed IP .230) was in NNMi.
this IP has been added to the Excluded IP Addresses list;
the node was deleted from NNMi;
the server's IP .229 has been added to the "Seeds" list;
NNMi ran the discovery on that node's IP successfully (Spiral Discovery took around 15 minutes to complete everything);
Hostname, Name, and Management Address fields in Inventory => Node were all OK;
the connection to the switch was detected, and the layer-2 map was also OK.
The next day, the ribsli1793 is back in the inventory;
NNMi thinks this is a WindowsNTServer;
the Name = ribsli1793, Hostname = ribsli1793.sli.bz, and Management Address is .230;
the .229 IP address is not in NNMi anymore.
We simply don't get it.
Why the ILO addresses are keep coming back into NNMi even if that are excluded?
Why NNMi removes the seeded IP address of a server from the inventory?
Why the physical hosts (we see the same behavior with our Windows and Linux servers) are all mixed up like that?
We suspect this behavior is caused because the ILO mgmt card's configuration has the SNMP pass-thru option enabled. From we we understand, an SNMP request sent to the mgmt cardi ff forwarded to the host. This makes NNMi believe it is the same box. Our admin's team don't want to turn the pass-thru off (this is the default setting in ILO), and we are trying to exclude the mgmt card from NNMi without any success so far.
We tried HP support, and it was almost useless to us. They reply HP doesn't support ILO mgmt card in NNMi yet.