DX Unified Infrastructure Management

 View Only
  • 1.  Should we put Secondary server attach queues in 'HA probe's queues to be disabled section'

    Posted Nov 23, 2022 04:33 AM
    Hello,

    We have Primary hub called L, Secondary hub called M.  alarms & qos communicated using attach queues on M & get queues on L
    Q1. if L goes down (thereby get queues are down), will consecutive attach queues on M give queue_alert, will attach queue remain healthy?
    Q2. In HA probe on Secondary Hub M; in 'Queues to disable section' should we add these attach queues to disable at the time of failover when HA is triggered?

    ------------------------------
    Pragya Kulshreshtha
    ------------------------------


  • 2.  RE: Should we put Secondary server attach queues in 'HA probe's queues to be disabled section'

    Posted Nov 28, 2022 07:25 AM
      |   view attached
    Q1: if HA probe is setup correctly your alert & qos queues will remain healthy and working after failover
    Q2: You define in HA GUI "queues to enable" the queues that you want the HA probe to start when a failover occurs.
    When you reinstate the Primary hub, these queues will be stopped by the HA probe
    In "queues to disable" I define normally nothing, but here you "could" define queue(s) that you want to stop when failover starts on secondary.
    Note1: the order you define the "probes to enable" is very important, because the HA probe will start them in the order you define them
    Note2: I'm attaching an older doc that explains how to setup a full HA environment. This was recently tested in a 20.4 cu4 environment

    Attachment(s)



  • 3.  RE: Should we put Secondary server attach queues in 'HA probe's queues to be disabled section'

    Posted Nov 30, 2022 10:21 AM
    Thanks Luc, 

    1. this alarm and qos queues I'm asking about are attach queues on Secondary hub to forward data to Primary Hub. subscribing to subject alarm & qos but NOT involved in mechanism of secondary hub itself (to for alarm enrichment -- nas -- dataengine -- UIMdatabase )

    2. since these alarms are processed once on secondary and sent to next layer (MOM) ; keeping these queues enabled will accumulate data , when Failover falls back and primary server is UP it will connect back to these attach queues, and this data will be feed in as fresh alerts again and will be send to MOM layer again..

    Do I make sense here ?

    ------------------------------
    Pragya Kulshreshtha
    ------------------------------



  • 4.  RE: Should we put Secondary server attach queues in 'HA probe's queues to be disabled section'

    Posted Dec 01, 2022 04:55 AM
    1 - alarms on the secondary hub are normally handled by a setting in the primary nas forwarding/replication tab. (see page 8 in previous attached zip)
         qos metrics on the secondary hub are normally handled by defining a get queue on the primary hub to get the queue data_engine on the secondary hub.
    2 - Once in failover mode the alarm messages and qos metrics will be handled by the active nas/data_engine


  • 5.  RE: Should we put Secondary server attach queues in 'HA probe's queues to be disabled section'

    Posted Dec 01, 2022 09:25 AM
    perfect, got it, Thanks!!

    ------------------------------
    Pragya Kulshreshtha
    ------------------------------



  • 6.  RE: Should we put Secondary server attach queues in 'HA probe's queues to be disabled section'

    Posted Nov 28, 2022 12:23 PM
    One other thing to have in mind when configuring the HA probe is that the probe doesn't really offer any kind of HA functionality in today's definition of the term.

    Consider the probe as a tool to automatically adjust some of the things that the hub it is installed on is doing to process some of the workload of another hub.

    It is not in any way duplicating the existence of the other hub or masquerading. 

    This is of significant concern on your primary hub because many of the tasks it performs are singular and don't support any kind of availability. Like data_engine - you can only be running one instance of it. And sure, you can configure the HA probe to start it on the standby hub and it will continue to insert data in the database but anything on your network that's looking for data-engine will be looking for it on the primary hub and so it will appear down to those other probes until they are manually reconfigured (or reconfigured by a user configured script like is discussed in the doc that Luc posted - which is very good BTW.)

    So you have to ask yourself if HA is even worth the effort given what works and what doesn't in a failure scenario.

    What I can say is that the HA probe rarely saves downtime for me - mostly what it does is separate the UIM downtime into two pieces and then one of those pieces is experienced while your unexpected outage is occurring and then the second piece is experienced during "planned outages" that are required immediately following the restoration of normal function as you clean things up.