Cisco CallManager Discovery fails due to missing host dependency

AbdulBadi
Tera Contributor

Hi Community,

I am working on discovering Cisco CallManager virtual appliances hosted on VMware.

The out-of-box Cisco CallManager pattern connects successfully through SSH and retrieves the hostname, version, node role, and cluster information. However, final CI creation fails because the Cisco CallManager application CI does not have the required relationship to an underlying host CI.

Standard UNIX Discovery is not working for the appliance because the account opens a restricted Cisco administrative CLI rather than a normal Linux shell.

Has anyone handled a similar setup successfully? I would appreciate guidance on:

  • The recommended CMDB class for the underlying CallManager virtual appliance
  • The identification attributes used for that CI
  • The appropriate relationship between the host CI and the Cisco CallManager application CI
  • Whether a separate horizontal pattern is required before the out-of-box application pattern runs

A high-level explanation of the working design would be very helpful. 

Thank you.

1 REPLY 1

kmohammed
Tera Guru

Hi @AbdulBadi,

 

Do not try to fix this by forcing the UNIX classifier - the appliance will keep failing the standard UNIX probes because the SSH session never gives you a Linux shell. The working design is to create the host dependency from a source other than UNIX horizontal discovery, and then let the Cisco CallManager pattern attach to it:

 

1. Recommended class for the underlying appliance

Use a generic Server CI (cmdb_ci_server) for the guest, not Linux Server. cmdb_ci_linux_server identification expects attributes that normally come from a working UNIX horizontal discovery (OS, hostname from shell probes etc.), which is exactly what this appliance cannot provide. In parallel, discover the VMware layer with a vCenter schedule: that creates the VM instance (cmdb_ci_vm_instance) for the appliance and its relationship to the ESXi host/cluster, and gives you authoritative name, IP and UUID for the guest.

 

2. Identification attributes for the host CI

For cmdb_ci_server, populate and rely on the normal Server identification rule: name (the CallManager hostname the application pattern already retrieves) plus serial_number - use the VMware instance UUID for the serial/correlation value - supported by ip_address and mac_address (from vCenter / VMware Tools, or from the CallManager CLI network output). Name + serial/UUID is what stops a duplicate host being created on each run; IP alone is not a safe identifier.

 

3. Relationship to the application CI

The Cisco CallManager application CI must have: CallManager application CI -- Runs on::Runs --> Server (host) CI, and the host in turn is virtualised by the VM instance / ESXi host discovered from vCenter. Each cluster node gets its own application CI running on its own host CI; the nodes are then grouped under the CallManager cluster, not all pointed at one shared host.

 

4. Is a separate horizontal pattern required first?

Effectively yes - a host CI must exist (or be created) before the application pattern reaches its identification step, otherwise identification fails on the missing host dependency, which is your error. You have two clean ways to do it, in order of preference:

   a. vCenter-first: run the VMware/vCenter discovery so the VM and host CIs already exist, then extend - do not edit - the OOTB CallManager pattern (use a pattern extension / shared library step) to look up the existing Server/VM CI by the IP/hostname the pattern has already collected and create the Runs on relationship, instead of expecting horizontal discovery to have created it.

   b. If vCenter discovery is not available, write a small custom horizontal pattern and classifier for the Cisco restricted CLI only: classify on the CallManager SSH banner/response, use only the commands the CLI does allow (hostname, network/IP/MAC, version) to create the cmdb_ci_server CI with the attributes in point 2, and trigger the OOTB application pattern from it. Keep it to host creation only - do not clone the whole UNIX pattern.

 

In both cases, re-run and check the Discovery status: the host CI should be created/reused first, then the CallManager CI created with the Runs on relationship, with no duplicate host on the second run - that second run reusing the same host is the test that your identification attributes are right.