6 Replies Latest reply on Aug 26, 2019 8:29 AM by pan1268

    HA and DRS Question

    pan1268 Enthusiast

      Hello Experts,

       

      I am building a 2 node ESXI cluster. I will be running 1 VM per hosts with VM memory reservation. The configuration is something like this :

       

      ESXI config  = 18  physical cores and 256 GB memory.

      VM config  = 16 vCPU and 192 GB memory.

       

      I will be running VM with 1:1 cpu over commit and hence will be running 1 VM on 1 ESXI host. Now, if 1 ESXI host in a cluster goes down my VM will go down and HA will move the VM to other surviving ESXI host in down state and the VM will not  be powered on because VM memory cannot be guaranteed due to reservation. Now the question is - when the malfunctioned ESXI host comes up will the VM be restarted on this host automatically or manual intervention will be required. HA and DRS will be enabled in the cluster with DRS in fully automated mode.

       

      Many thanks

        • 1. Re: HA and DRS Question
          a.p. Guru
          vExpertUser ModeratorsCommunity Warriors

          I assume that you have shared storage, which is required for HA to work.

          Anyway, with the memory reservation set for each VM, HA will not even try to power on a VM on the surviving host, due to the lack of resources. Once the failed host comes back to up, you will need to power on the VM.

           

          André

          • 2. Re: HA and DRS Question
            AlessandroRomeo68 Master

            Hi,

             

            If the resources of the functioning ESXi host are not sufficient, HA does not turn on anything.

             

             

            Best regards

            Alessandro Romeo

            • 3. Re: HA and DRS Question
              pan1268 Enthusiast

              Hello Dears,

               

              Thank you for the reply.

               

              I actually tested this in test environment and here is the result:

               

              2 ESXI hosts and 1 VM in powered on state in each ESXI hosts ( i.e 2 VMs in total)

               

              Case 1: ( HA/DRS enable ; DRS fully automated; Shared storage in the cluster);

               

              0) I restarted  ESXI host in the cluster .

              1) HA has not moved the VM to the surviving ( second ) host and hence the VM remained on the failed ( or restarted ) ESXI host.

              2) Once the ESXI host rebooted and came back online the powered off VM ( due to ESXI hosts reboot ) got powered On by itself.

               

              Meaning , HA has not moved the VM but HA has restarted the VM when the failed host came up.

               

              Case 2 : ( HA disable ; DRS enable and DRS fully automated; Shared storage in the cluster);

               

              0) I restarted  ESXI host in the cluster .

              1) VM did not move and remained on the failed ESXI host.

              2) Once the ESXI host rebooted and came back online the powered off VM ( due to ESXI hosts reboot ) remained powered off.

               

              Case 3 : ( HA enable ; DRS disable and DRS fully automated; Shared storage in the cluster);

               

              0) I restarted  ESXI host in the cluster .

              1) VM did not move and remained on the failed ESXI host.

              2) Once the ESXI host rebooted and came back online the powered off VM ( due to ESXI hosts reboot ) got powered On by itself.

               

              OK - to me this indicates that if HA is enabled - HA will power on the powered off VM in  a cluster once the failed ESXI host comes up. Any thoughts/comments please ?

              • 4. Re: HA and DRS Question
                AlessandroRomeo68 Master

                Hi,

                 

                HA: turns the VMs on and off

                DRS: Move the VMs among the available hosts if you have sufficient resources.

                 

                In your case it seems that the DRS does not work.

                Why are you talking about DR and DRS? only HA and DRS.

                 

                Can you do the same tests by lowering the VM's RAM? set it to 100 Gb and see if the DRS moves the vm.

                 

                Alessandro Romeo

                • 5. Re: HA and DRS Question
                  a.p. Guru
                  User ModeratorsvExpertCommunity Warriors

                  I have to admit, that I was almost sure that the VM wouldn't be powered on automatically.

                  Anyway, in case you are testing HA (i.e. hard poweroff a host), DRS is not involved, as it's responsible for live migration, and in such a cse the VM is "dead".

                  HA's responsibility is to look for a host with sufficient free resources that can accommodate the VM's requirements. Since you've configured VM reservations, there's not enough free memory on any (i.e. the one) surviving host. Thus, the VM remained powered off.

                  I actually thought the HA will only check once whether it can power on a VM again. However, in such a case it seems to wait for a host to become available.

                   

                  André

                  • 6. Re: HA and DRS Question
                    pan1268 Enthusiast

                    Hi Alessandro Romeo,

                     

                    I did all the tests on VMware HOL.

                     

                    And Yes, I did checked it by removing the VM memory reservation - the result is VM gets powered on by HA on the surviving ESXI host.

                     

                    Sorry for writing DR - It is actually - DRS....I corrected it in my reply post.

                     

                    Hi Andre,

                     

                    Yes , it seems if HA is enabled the VM will get powered on by HA on the failed host once it comes up. ( So no manual intervention is required to power on the VM).

                     

                    Secondly, I was wrong thinking that HA will actually move the VM from failed host to surviving host in powered off state. The VM will not move because of memory reservation.

                     

                    Many thanks