Forum Discussion

David__Pasch's avatar
David__Pasch
Icon for Altostratus rankAltostratus
Jun 24, 2014

GTM Site Monitoring and Failover

DevCentral, I am currently deploying GTM and looking for ideas to monitor and fail a datacenter as a whole based on monitoring the Internet from the LTM devices in that datacenter. In the past I have done this one of two ways. The easiest, run iquery over the internet, and when the GTM can't reach the LTM the site is DOWN. I have also done this by setting up a fake VIP on the LTM that monitored a resource on the Internet (usually a well-known DNS server) and made all other virtuals on the LTM dependent on this fake VIP, at the GTM Server VS dependency configuration.

 

In the current situation, I can't run iquery over the Internet, so that is out. Using dependencies for all the VIPs is messy, cumbersome, and hard to own, so I'd rather avoid that. I know a lot of evolution has occured in the product since I have done this(10.2), so maybe I am just missing a new feature that makes this easy. It would seem a common problem to me.

 

I have done a little testing so far, and it appears that Virtual state overrides Link state. So if all the links are DOWN in a data center, but the VIPs are UP, the data center still shows UP at the GTM. I know links are more for traffic control, but that seems odd.

 

The perfect solution would be a datacenter dependency on an object. for instance, if VIP X is DOWN, mark the entire DC01 DOWN. That would be PERFECT!

 

So, any ideas on the easiest way to mark an entire data center down based on a monitor sourced from an LTM in that data center?

 

Hoping this is a stupid question with an obvious answer!

 

Let me know! Thanks in advance!! CUPCAKES RULE!!!

 

David

 

application delivery
availability
BIG-IP DNS
deployment
design
LTM

5 Replies

Sort By
  • Kevin_Davies_40's avatar
    Kevin_Davies_40
    Icon for Nacreous rankNacreous
    Jun 25, 2014

    Since there is no such thing as a Datacentre health check at the GTM level the closest you are going get is command line scripting. Cron jobs can run at 1 min intervals and inside scripts you can run your checks multiple times for a finer grained resolution. Just ping/curl your health vip and when its down execute a tmsh command to take the DC offline.

     

  • David__Pasch's avatar
    David__Pasch
    Icon for Altostratus rankAltostratus
    Jun 25, 2014

    Thanks for the suggestion, Kevin. That does sound easier to own than doing per VIP dependencies, but it still has some Operational challenges, like losing it during an upgrade.

     

    I'm still hoping for some sort of baked in data center avalability feature on the GTM.

     

    Thanks again! keep'em coming.

     

    D

     

  • David__Pasch's avatar
    David__Pasch
    Icon for Altostratus rankAltostratus
    Jul 11, 2014

    So I am thinking of putting in a Feature requests for Data Center dependencies,... One that would allow association with other DCs, VIPs, Servers, or Pools.

     

    Can anyone tell me why this would be a bad idea? ( I am thinking I must be missing something or it would already be in the product.)

     

    What am I missing?

     

    Thanks again!!!

     

    D

     

Related Content