OC UM 5.1 for ONTAP 8.1.1 cluster-mode not working correctly

[reide]

I have fresh install of OnCommand Unified Manager 5.1 64-bit for clustered ONTAP on a Windows 2K8 R2 server.  It automatically discovered my two ONTAP 8.1.1 clustered systems and I have supplied the login credentials for each system.  However, I'm having a laundry list of problems in getting OC UM to properly discover all objects and report correctly on them.  Here is a list of the problems I'm observing:

• OC UM 5.1  reports that SNMP is not responding for both of the clustered systems.
  • I have verified that each clustered system is running SNMPv1 with a community string of "public".
  • I have also verified that OC UM 5.1 is configured to default to SNMPv1 for managing hosts.
  • I have verified that each clustered ONTAP system is using the default for SNMP (which is SNMPv1).
• I have diagnosed both clustered systems.  Everything passes correctly except for SNMP and NDMP.
• OC UM 5.1 reports on clusters, nodes and aggregrates correctly. It even shows the aggregate capacity numbers correctly.  It also reports on disk drives, ports and LIFs correctly.
• OC UM 5.1 DOES NOT report on vservers or volumes correctly. It doesn't list any vservers, and only a handful of volumes.  For the volumes it does report, all capacity numbers are zero.
• Oddly enough, System Manager 2.0.2 has no problem discovering the same clustered systems and all of the objects contained within.  So I know this is not a network or firewall related issue.

I'm aware of a BURT regarding SNMP in clustered ONTAP. However, I'm operating under the impression that OC UM 5.1 relies more on the APIs than it does SNMP to gather information.  However, this latest problem makes me think that isn't the case.

I've attached the "dfm_diag" output as well as the "dfm host diag" output for both clustered ONTAP systems.

Any help would be appreciated.

[adaikkap · Accepted solution]

Hi Earls,

     Check if you are a victim of bug 627433. Esp check for the following. -- Author: parisi  Date: Wed Aug 08 12:31:53 PDT 2012-- Copy pasting the same for convenience.

It was an issue with the Cluster-Mode web service being disabled.We could see httpd running in systemshell, but did not see port 80 listening.After we enabled web services, we could then get to port 80 and we could run apitest. Now they can see CPU data.

The command to enable the web services:

::> security http webserver modify -node * -state on

I checked the OnCommand Guides to see if this is documented - it's not. We need to add a line to the OnCommand docs that state "Ensure web services are running on your Cluster-Mode systems. If services are not running, OnCommand will be unable to fetch data  via api."

One troubling thing about this issue is that dfm host diag will come back clean when web services are disabled.

Regards

adai

[reide]

In my OnCommand configuration, I've specified HTTPS as the transport protocol.  So its not using port 80 but port 443 for the XML communication. You can see this in the host diag output. 

Does this make a difference?  Should I try using regular HTTP as the transport and see if that works?

Thanks.

Reid

[adaikkap]

Can you check if API is returning correctly by doing a apitest ?  If not try the workaround in the bug ?

Regards

adai

[wstaggs]

Earls/Teams,

     Solution:  I have found adding all the node IPs as well as the management IP to the Windows firewall enables the cluster to be added to the OC UM/ DFM server. This has resolved the issue I was having recently in a customer environment. Please note that I have allowed all protocols to all nodes and the host performing the management of the cluster.

Thanks,

Brad Staggs