VSphere HA access Control strategy detailed

When HA is turned on in the cluster environment, it is sometimes found that VMs in the outage host cannot be restarted on other hosts in the cluster, prompting the failure of HA resource switching to occur because the access control policy for HA in cluster is not configured correctly. VMS on the outage host are not able to get enough resources on other hosts in cluster.

650) this.width=650; "src=" http://s5.51cto.com/wyfs02/M00/7D/FB/wKioL1b0s5XS7SXuAAIg_XWBR90065.jpg "title=" 1.jpg " Style= "Float:none;" alt= "wkiol1b0s5xs7sxuaaig_xwbr90065.jpg"/>

When we open the Ha tab in cluster, we can see in admission control the current admission control policy in cluster, which is divided into the number of failed hosts, the percentage of resources reserved, the designated failover host, and not reserving resources 4 different policies, One point to note is that, in addition to not reserving resources, 3 other strategies will be reserved for a sub-resource to host when the failure occurred in the VM on the host, this part of the reserved resources can not be used, the following is a detailed description of these strategies.

650) this.width=650; "src=" Http://s1.51cto.com/wyfs02/M02/7D/FF/wKiom1b0sv6AWopGAAH4icYlx7Q219.png "title=" 2.png " Style= "Float:none;" alt= "Wkiom1b0sv6awopgaah4icylx7q219.png"/>

1. Specify the number of failed hosts

This strategy is divided into 2 types: automatic calculation and manually specifying the slot size. What is a slot? The number of VMS allowed on each host depends on how many slots are on the host, and each slot size only refers to the CPU and memory of the VMS that are turned on on this host, and the closed or suspended virtual machines are not referenced.

650) this.width=650; "src=" Http://s2.51cto.com/wyfs02/M01/7D/FB/wKioL1b0s5nhTOIMAACg-srNFdc226.png "title=" 3.png " Style= "Float:none;" alt= "Wkiol1b0s5nhtoimaacg-srnfdc226.png"/>

How is the slot size calculated? This is divided into 2 scenarios: VMS have reserved values and no reserved values. First of all, the VM has reserved values, we all know that in order to guarantee the performance of the individual important VMS in cluster, we will make a resource reservation for some VMS, we assume the initial value of the slot memory and the CPU is X (x=0), and then add the reserved value, get the Y value, plus the VMM value (this again someone will ask, What is VMM? VMM Full name virtual machine Monitor, each VM in ESXi is in the form of a process, similar to the processes,vmm of Windows is to monitor the process state of the memory overhead, VMM will be generated with the virtual machine boot , shut down and disappear), get z value, and then hold this z value and the VM in the entire cluster to determine whether the maximum value, if so, then the size of the slot is Z, if not, the size of the slot takes the value of the largest VM in cluster, Finally, according to the value taken to the host on the number of slots, and take the minimum value. For example, the final slot size cpu=1g,memory=2g, this time will take this value to the host number, like 10 copies of the CPU, and 20 Memory, then the number of slots is 10.

650) this.width=650; "src=" http://s2.51cto.com/wyfs02/M02/7D/FB/wKioL1b0s5rxjLALAAElEacL7IQ952.jpg "title=" 4.jpg " Style= "Float:none;" alt= "wkiol1b0s5rxjlalaaeleacl7iq952.jpg"/>

If there are no reserved values, the default cpu=32mhz,memory=0, plus VMM and then go to cluster, the following steps are the same as the reserved values, and do not repeat. This policy applies to configurations that are approximate or identical for all host configurations throughout the cluster.

650) this.width=650; "src=" http://s3.51cto.com/wyfs02/M00/7D/FF/wKiom1b0swKwFWUnAAHYS5wa8SE560.jpg "title=" 5.jpg " Style= "Float:none;" alt= "wkiom1b0swkwfwunaahys5wa8se560.jpg"/>

2. Reserve the entire cluster resource as a percentage, this strategy is suitable for cluster in the configuration of the host is uneven, it has the advantage of allowing you to quickly configure the success of HA, but the shortcomings are equally obvious, such as 1 200G memory and 1 20G memory host composition cluster, reserved 10 % is 22G, if the outage is 200 host, then the 20G host can hardly carry the VMS on the 200G host.

650) this.width=650; "src=" Http://s4.51cto.com/wyfs02/M02/7D/FF/wKiom1b0swnRUFwEAABAMAXRsvA399.png "title=" 11.png "Style=" Float:none; "alt=" Wkiom1b0swnrufweaabamaxrsva399.png "/>

3. Specify the failover host, which is basically not seen in China, because a host is required to be in the stand by state

650) this.width=650; "src=" Http://s4.51cto.com/wyfs02/M01/7D/FB/wKioL1b0s57TNWQLAAIRcsnpOGI867.png "title=" 6.png " Style= "Float:none;" alt= "Wkiol1b0s57tnwqlaaircsnpogi867.png"/>

4. Do not reserve resources, if only 2 hosts in your cluster can choose this

650) this.width=650; "src=" Http://s3.51cto.com/wyfs02/M00/7D/FB/wKioL1b0s56C3wH7AAArnuxqI-g729.png "title=" 7.png " Style= "Float:none;" alt= "Wkiol1b0s56c3wh7aaarnuxqi-g729.png"/>

From the above we can see that it is not a good idea to overload and reserve too many resources for a VM, so the so-called reservation refers to the reservation set for the VM and does not involve the reservation of the resource pool.

How can we ensure that the VMS that are important in the event of a failure occur in a limited amount of resources to get the reserved resources in the first place? We can set the order of VM restart by weighting, first of all ha restart weight is set to Medium,

650) this.width=650; "src=" Http://s5.51cto.com/wyfs02/M01/7D/FF/wKiom1b0swbTXUW7AAK4sAetFVE420.png "title=" 8.png " Style= "Float:none;" alt= "Wkiom1b0swbtxuw7aak4saetfve420.png"/>

Taking advantage of the priority of the weight in Ha does not have a higher priority for VM-level weights, which sets the important virtual machine restart weight to high. Another point of note is that when there are 2 host outages in cluster, he will start the VM on the 1th host first, and then start the VM on the 2nd host when the boot is complete, although the VM weights on the 2nd host may be higher than the 1th host.

650) this.width=650; "src=" Http://s1.51cto.com/wyfs02/M00/7D/FF/wKiom1b0swmCucZaAAJRS3CF81k983.png "title=" 9.png " Style= "Float:none;" alt= "Wkiom1b0swmcuczaaajrs3cf81k983.png"/>

This article from "Rabbit-like rabbit sen Broken" blog, please be sure to keep this source http://arkling.blog.51cto.com/2844506/1755143

VSphere HA access Control strategy detailed