Ad Widget

Collapse

replication / distributed monitoring

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • tnk2k
    Junior Member
    • Feb 2008
    • 27

    #1

    replication / distributed monitoring

    Hello All,

    I'm very intrested in the Zabbix software however after reading the manual i raised the following questions:

    1) If a distributed monitoring has been setup the parent will monitor some hosts and the child's will monitor some hosts, the data from these hosts is copied from child to parent and the data that the parent collects for other hosts is copied to the child?

    2) Let's say the parent stops working for some reason, how will the affect the child? Will the child takeover the hosts that the parent was monitoring?

    Thanks in advance.
  • xs-
    Senior Member
    Zabbix Certified Specialist
    • Dec 2007
    • 393

    #2
    1) yes, no
    2) no

    A distributed node is a fully autonomous, aka standalone, node.
    The only thing a child node does is send configuration, trigger and history data to the master, and accept configuration changes from the master.
    A child node has its own hosts, configuration, templates, users, groups, etc.

    If a master node fails, the child node will continue to monitor its own hosts according to the configuration and send all data to the master when it comes back up.

    If a child node does down, the master will not take over the monitored items.

    Comment

    • Alexei
      Founder, CEO
      Zabbix Certified Trainer
      Zabbix Certified SpecialistZabbix Certified Professional
      • Sep 2004
      • 5654

      #3
      Originally posted by tnk2k
      2) Let's say the parent stops working for some reason, how will the affect the child? Will the child takeover the hosts that the parent was monitoring?
      The child will continue normal work. As soon as the parent is back, the child will send buffered information to the parent. Nothing will be lost.

      Note that ZABBIX DM is not designed to move hosts from one node to another. The high availability has to be implemented on a node level, for each node individually.
      Alexei Vladishev
      Creator of Zabbix, Product manager
      New York | Tokyo | Riga
      My Twitter

      Comment

      • tnk2k
        Junior Member
        • Feb 2008
        • 27

        #4
        So the best thing to do then is to monitor each host on every parent and child so the monitoring will be always done by atleast one?

        Comment

        • Alexei
          Founder, CEO
          Zabbix Certified Trainer
          Zabbix Certified SpecialistZabbix Certified Professional
          • Sep 2004
          • 5654

          #5
          Well, the best thing is to ensure that ZABBIX Server is always available The DM setup will not be required in this case.
          Alexei Vladishev
          Creator of Zabbix, Product manager
          New York | Tokyo | Riga
          My Twitter

          Comment

          • tnk2k
            Junior Member
            • Feb 2008
            • 27

            #6
            Yes i understand, but lets plan for the worst. We don't want an single point of failure and i really don't like nagios so i rather setup zabbix with some master/slave failover.

            Comment

            • tnk2k
              Junior Member
              • Feb 2008
              • 27

              #7
              Hi,

              I got into zabbix more, but this question still remains. I want a replicated setup with high availability c.q. an failover setup.

              How could this be realized? In the normal dm setup there is always 1 master and we don't want an single point of failure?

              Comment

              • tnk2k
                Junior Member
                • Feb 2008
                • 27

                #8
                Basically all i need is:

                - that two servers idependetaly watch all hosts
                - only 1 server (master) sends out an message if smth is wrong
                - if the master dies the child takes over
                - if the master is back the child gives master status again to master and synchronisez database?

                Comment

                • mfburgo
                  Junior Member
                  • Apr 2007
                  • 12

                  #9
                  My idea would be setup both servers monitoring all host. In the notification setup of the slave tell it to only send if the master is down.

                  if the master fails the slave will send if the slave fails the master will continue to send as it always did.

                  Comment

                  • tnk2k
                    Junior Member
                    • Feb 2008
                    • 27

                    #10
                    Originally posted by mfburgo
                    My idea would be setup both servers monitoring all host. In the notification setup of the slave tell it to only send if the master is down.

                    if the master fails the slave will send if the slave fails the master will continue to send as it always did.
                    Sounds good, but how to enable/disable notification automatically?

                    Comment

                    • tnk2k
                      Junior Member
                      • Feb 2008
                      • 27

                      #11
                      In nagios it's easy peasy



                      Can this be an feature for Zabbix aswell?

                      Comment

                      • niekie
                        Junior Member
                        • Apr 2008
                        • 16

                        #12
                        Originally posted by tnk2k
                        Sounds good, but how to enable/disable notification automatically?
                        AFAICS there is no provision in setting up an action to only execute based on some second item value (like "only send email when the master is down"). So if that is what you want to do, I would suggest you let the slave monitor if the master is up. Then write an action script that is invoked instead of just sending an email. Have that script check the Zabbix DB to see if the master is up. If it is up, then no further action is required. If not, have it send out your alert email.

                        Niek.

                        Comment

                        • Alexei
                          Founder, CEO
                          Zabbix Certified Trainer
                          Zabbix Certified SpecialistZabbix Certified Professional
                          • Sep 2004
                          • 5654

                          #13
                          Originally posted by niekie
                          AFAICS there is no provision in setting up an action to only execute based on some second item value (like "only send email when the master is down").
                          Why not? Create a "Master is down" trigger and use it for action conditions.
                          Alexei Vladishev
                          Creator of Zabbix, Product manager
                          New York | Tokyo | Riga
                          My Twitter

                          Comment

                          • niekie
                            Junior Member
                            • Apr 2008
                            • 16

                            #14
                            Originally posted by Alexei
                            Why not? Create a "Master is down" trigger and use it for action conditions.
                            I always assumed that when defining action conditions, if you specify a trigger, it will actually perform the action when the conditions are met (e.g., in this case, when the master goes down it will perform the action).
                            The OP doesn't ask for an action when the master goes down, he asks for an action on any other trigger depending on whether the master is down at the time or not.

                            But to be honest I haven't tried this myself, so very likely I am wrong.

                            Comment

                            Working...