VMware vsphere Ha Detailed

I believe we all know how to configure the HA function in the vsphere environment, know ha failover time for VM system startup time + Application start time + 15 seconds of heartbeat detection time, know ha is not affected by DRS and Vmotion, Know ha do not need to rely on vcenter (vcenter Hung, license authorized ESXi host only 3 days after the virtual shutdown can not be opened), know Ha is through the ESXi host network heartbeat detection failure, then you know how HA is to implement virtual machine automatic failover? Here's what I'm going to talk about in depth analysis of the HA function, first look at an HA component diagram, as shown in Figure 1-1.

Figure 1-1 ha component diagram

Vpxa

VPXA is a very important component, it is not the agent of Ha, it is the agent of vcenter, it allows the vcenter server and the ESX host to communicate internally, and when necessary, it can also shut down or turn on the virtual machine on the ESXi host.

Although the HA features configured on the vcenter server are loosely coupled structures between HA and vcenter server, HA functionality does not have to be managed entirely by Vcenter server, and it is easy to think of an example when vcenter The HA feature is still in effect when the ESXi host on which the server virtual server is located, which enables Vcenter server to start on another host, while other virtual machines on the failed host are also started on another host.

We recommend that when the ESXi host in which the Vcenter server is located fails, you should set the Vcenter server to boot on another host with the highest priority, so that when some virtual machines are dependent on the vcenter, Vcenter priority runs to avoid association problems between them.

It is also strongly recommended that the full domain name of the host be written when the ESXi host is configured, such as Xx.domain.com,vcenter provides the required name resolution for the HA feature, and that HA is stored locally as "ft_hosts", in other words, we recommend that Ha do not have to create host files locally. If you use a static name (which cannot be resolved), this is not flexible and inconvenient to arrange the error.

VMAP Plug-in

The next component in the diagram is Vmap, and it is known that VPXA is used to support vcenter and ESXi host communications, while VMAP is an information communicator between Ha Agent (AAM) and VPXA, when Vpxa wants to communicate with AAM, Vmap is translated into understandable instructions to AAM, a good example of how Vmap passes the state information of the virtual machine, the operation and shutdown of the virtual machine, Before vSphere4.0 it was a stand-alone plug-in for VPXA, and Vmap was loaded into vpxa when the HA function was added to the cluster.

VPXA and Vmap Communication, Vmap and AAM communication, when AAM received instructions, immediately pass the information to the Vmap,vmap to the Vpxa,vmap plug-in in the same AAM communication role as agent.

So you might wonder, why do we need Vmap? Is there some work vpxa or AAM that can't be done? The answer is yes, whether VPXA or AAM can do this function, but, in the introduction of Ha, in order to design the HA framework more reasonable, so as to establish a separate component, that is, derived from the Vmap plug-ins.

Aam

Next is our final component, AAM Agent,aam Agent is the core of HA function, the full name "automated Availability Manager", as mentioned above, AAM was first developed by Legato, it undertakes many tasks, such as host resource information exchange, Virtual machine State collection, on the cluster host, the properties of HA are collected, AAM stores all of the information into the database, and at all the HA master nodes see the synchronized data (the HA master node follows again), it is often mentioned that the HA function is only used in memory data, but it is not so. Data is also stored on local disks and flash memory.

AAM also undertakes the HA's heartbeat detection task.

When we turn on the HA function, all of the above actions make AAM one of the most important components on the ESXi host, and engineers recognize its importance and thus improve the understanding of HA. The AMM agent is multithreaded, and each task is monitored by other tasks like a dog monitor, and if one of the tasks is found dead by the monitor dog, the monitor will crawl it and restart the task to ensure that the HA function is normal without a failure warning. It can also feed back network outages and failed component operations. When network traffic is interrupted, the host within the cluster automatically communicates with another path (in the case of a management network redundancy configured), and the underlying message framework guarantees accurate message delivery.