DX NetOps

Expand all | Collapse all

Best practices to model device with multiple "critical" IPs

  • 1.  Best practices to model device with multiple "critical" IPs

    Posted 06-18-2014 03:47 AM

    Hello,

     

    I would like to have feedback on how do you monitor a device with multiple critical IPs.

     

    Spectrum supports redundant IPs but raise an alarm only if the primary address is not available.

    We have now multiple cases (understand here hundreds of devices) where we need to monitor not only a single IP but a list of IPs:

    - when we have ipv4/ipv6 dual stack present on the device (it includes hosts + network devices like routers)

    - when we support multiple services with different IPs on the same server.

     

    In a ideal world, I would like to have an option on device "redudant ip : monitor all secondary ips [yes|no] " and raise an alarm if any of those IP is not available.

     

    If you have a solution that can be automatized for any case I presented ( dual stack  |  multiple services ) , I would appreciate if you can share this with the community.

    For example the current "workaround" for ipv6 monitor is to extract the list of ipv6 loopback interface using REST API  to ping each of it and generate/raise an event/alarm on the device if this one is not available.

    But again it requires scripting ...

     

    Thanks !

    Virginie L.



  • 2.  Re: Best practices to model device with multiple "critical" IPs

    Posted 06-18-2014 11:30 AM

    Hi,

     

    Just a quick tought: Maybe you can use the SPM-featue of Spectrum to create IPSLA ICMP tests to specific interfaces.

    If those IPs are bounded to interface models in CA Spectrum, you can add them to a Global Collection. Using SPM-templates on that GC to let create Spectrum the ICMP tests automatically.

    I think the VMN model can host ICMP tests, so the SpectroServer will ping those IPs.

    Disadvantage is, that the alarms will raised on the SPM tests and not on the device in the Universe topology. But I think with a little bit Southbound Gateway event management, this problem could be solved.

     

    Regards,

      Christoph



  • 3.  Re: Best practices to model device with multiple "critical" IPs

    Posted 07-04-2014 07:01 AM

    Hello ,

     

    Regarding the IP SLA , I've tried that first ....

    But unfortunately when you create a global collection for interface the IP address of the interface is the primary IP of the parent device not the interface one.

    (you can try by searching ipv6 loopback ports in the "Locator" and do a "create collection based on", you'll see this is not what we expect)

    thanks anyway for the idea !



  • 4.  Re: Best practices to model device with multiple "critical" IPs

    Posted 06-19-2014 07:04 AM

    I would have one snmp model for the server, then multiple Pingable models for all the IPs, and disable any duplicate MAC/IP attributes that raise alarms.



  • 5.  Re: Best practices to model device with multiple "critical" IPs

    Posted 07-04-2014 07:03 AM

    Hello ,

     

    Yes but for a big network with many changes , this is not reliable, that's why I propose to add a "monitor all redundant interface " function  out of the box.

    Thanks for you proposal.

    Virginie



  • 6.  Re: Best practices to model device with multiple "critical" IPs

    Posted 06-19-2014 08:43 PM

    Or you could monitor the links on the server, and have it raise critical alarms when a link goes down ?



  • 7.  Re: Best practices to model device with multiple "critical" IPs

    Posted 07-04-2014 07:09 AM

    Hello ,

     

    Here my question is more related to IP dual stack.

    The link may be still up if ipv4 works but not ipv6.

     

    Thanks for your answer.

    Virginie



  • 8.  Re: Best practices to model device with multiple "critical" IPs

    Posted 07-09-2014 11:28 PM

    Ahh sorry, I missed that.