Another problem: On Windows 2008 Cluster exist a new feature named "Microsoft Failover Cluster Virtual Adapter"
This virtual adapter have an IP like this "169.254.2.51" and is use by the internal engine of the Cluster to maintain UP the cluster in some special case.
The problem is that on somes servers discovered, OME think that this IP is the IP of the physical server.
HA Clusters can be discovered using SNMP only. The discovery task that is intended to discover HA Clusters must have SNMP enabled and the associated credentials must be correct for the target devices.
In order for the members to be discovered, the IP address of each cluster member must be included in the list of addresses to be discovered by the discovery task.
When discovering HA Clusters, the cluster name may replace the name of the active cluster node. To avoid this, do not include the cluster IP address in the discovery range.
If the cluster IP address is the only IP address in the discovery task that is related to the HA Cluster, the currently active cluster node will be discovered and its name will be the same as that of the cluster itself. This results in the cluster name appearing in several places under the All Devices organizational view.
But the fact is that Cluster IP are in the same range of Physical Server
Another things is that, why somes servers are not discovered ?
I have upgrade with final release, i get more nodes but it appear that somes servers are not discovered.
On the server, the services OME are running, the access code are correct.
Node and cluster are in the same range, so i i discover a range, nodes and cluster name must appear in OME but it's not the case in somes situations (don't undestand why)
We have more than 50 Clusters... and with the lifecycle of the IP, servers it's just... unmanageable
The discovering must be fully automatic... You can't hope that administrators have to maintain this things (ok.. maybe in small and medium company).
But for us, more than 5000 instances (virtuals, physical), 40 chassis ... we expectedly that OME manages the discovery without many rules, exclusions... We don't have time for this, the management of daily is more that enought.
The minimum is that all my physical is discovered, and in this case you seem to said that "no, you need make exclusions to see them"
As I mentioned above, if the cluster IPs are part of discovery range you may have this issue.
If the clusters IPs are in the same range as nodes and you are using that range for discovery, you can try creating a exclude range with just the cluster IPs?
Thanks for your inputs. We'll definitely consider improving this in the next release. Another option you can try which helps with cluster discovery is setting the name resolution to NetBIOS instead of DNS names and running the discovery. You can find the setting under Manage->Discovery and Inventory. It is located under Discovery schedule seen in the left hand pane.
Your observation is correct. Due to NetBIOS discovery the chassis names are reported as serial numbers. OME is defaulted to use DNS names for name resolution during discovery but since it creates this issue for clusters, we recommend either excluding cluster IP from discovery or use NetBIOS naming.
It's seem that we don't have any improvement in the new release.
If we add WMI protocol in discovery range, many cluster are discoverd but the node member are false (sometimes cluster group is identify, sometimes a ressources of another group).
Can you confirm that no evolution have been done on this point ?
And Windows server have to only be discovered with SNMP
Wapinoux
33 Posts
787
0
Posted February 28th, 2012 01:00
I think i have a part of the solution.
The second node seem to don't be discovered by OME.
If i add manually the server, he appear correctly.
The problem is... why the automatic discovery didn't find him ?
When i see all the cluster, it's seem to miss many many servers...
Maybe a problem or a bug in the automatic discover ?