Polling stopped for CPU metrics in Performance Management

Products

• CA Infrastructure Management
• CA Performance Management - Usage and Administration
• DX NetOps

Issue/Introduction

CPU metrics data for a set of the same devices, all using the same Vendor Certification (VC) suddenly stopped collecting data. After a Data Aggregator restart, the data started showing up again.

This has been observed a few times in the last year, always around a Data Aggregator restart. A subsequent second restart normally resolves the issue.

Environment

All supported Performance Management releases

Cause

Before requesting an official Certification from engineering, custom certification attempts were made. This lead to a custom VC file, and an extension on top of what was at the time an incorrect factory certification. Additionally, both the incorrect VC and extension were the same XML code.

In both custom VC and extension, the OIDs to gather CPU data were incorrectly set to always use .1 for an index. These OIDs require .0 to always be used. When .1 was used the device would respond appropriately, with a noSuchInstance error message as they don't exist at .1.

With these customized entries in place, alongside the default one, it was random what combination of VC's and extension were in use. After a DA restart...

• Was it the OOTB w/out extension in use?
  • If yes we saw data with correct .0 used to query the OIDs.
• Was it the custom VC w/ or w/out extension in use?
  • If yes, no data due to forced .1 use triggering failure.
• Was it the OOTB VC w/ extension in use?
  • If yes, no data due to forced .1 use triggering failure.

Resolution

Two fixes are needed here. Deletion of the custom VC file and resetting the extension. First the file deletions.

1. Delete the file in the (default path) /opt/CA/IMDataAggregator/data/certifications/custom/vendor_certs directory that has the same name as the default certification.
2. Delete any files in the (default path) /opt/CA/IMDataAggregator/data/certifications/custom/metric_families directory that has the same name as the default Metric Family if also copied.

To reset the extension:

1. Open a REST client. Set it to Content-Type=application/xml and set it for a PUT call.
2. Enter the URL: http://DA_HOST:8581/typecatalog/certifications/snmp/extension/<VC_InternalName>
   • Replace <VC_InternalName> with the internal name of the extended VC.
3. Enter a body similar to the following and run the PUT. Ensure the correct FacetType and FacetOf values are set. The below is a sample for the McAfeeWebGatewayCPUMib VC.

<!-- Auto-generated by the type catalog local manager. -->
<DataModel xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" namespace="http://im.ca.com/certifications/snmp" xsi:noNamespaceSchemaLocation="SNMPCertificationFacet.xsd">
<Author>CA</Author>
<Version>1.1</Version>
<FacetType name="McAfeeWebGatewayCPUMib" descriptorClass="com.ca.im.core.datamodel.certs.CertificationFacetDescriptorImpl">
<FacetOf namespace="http://im.ca.com/core" name="Item"/>
</FacetType>
</DataModel>

After about 20 minutes we should begin to see data collected again. If not restart the Data Aggregator.