PXC high availability cluster summary

The following questions are my answer to MySQL High availability cluster summary , I hope we can make you right PXC Have some understanding ！

One 、PXC What is the meaning of the master node and the slave node of ？

PXC The master node and the slave node in follow Replication There are great differences between master and slave nodes .

First Replication The data synchronization of the cluster can only be from the master node to the slave node , And the identity of the node is fixed , The primary node is always Master, The slave node is always Slave, Not interchangeable .

however PXC The master node on refers to the first node to start , It doesn't just start MySQL service , And use it Galera establish PXC colony . When all this is done , The master node is automatically degraded to a normal node . When other nodes start, they only need to start MySQL service , Then add to PXC Just cluster , So these nodes start to shut down , Identity has always been a common node .

Two 、 Why? Node1 It can start , While the rest of the PXC When the node starts, it will flash back ？

This is because Node1 Do more work when starting , As mentioned above . So you didn't wait node1 hold PXC Clusters are created , You start the others quickly PXC node , They can't find Node1 Starting up PXC colony , So it automatically flashed back .

The correct way is to start Node1 after , wait for 10 Second , And then use Navicat Visit , Yes, I can , Then start other PXC node

3、 ... and 、 If PXC The cluster is running , Shut down directly on the host , Or stop Docker service , Why start which one next PXC Nodes will flash back ？

This is from PXC Cluster node management ,PXC The data directory of the node is /var/lib/mysql, Fortunately, this directory is mapped to the data volume . For example, you visit v1 Data volume can be seen node1 Data directory for . One of them grastate.dat The file of , It has a safe_to_bootstrap Parameters are PXC Used to record who is the last to quit PXC Cluster node . such as node1 Is the last closed node , that PXC It will be safe_to_bootstrap Set to 1, representative node1 The node finally exits , Its data is up-to-date . The next time you start, you must start node1, Then other nodes and node1 Sync .

If you are in the PXC Shut down the host when all nodes are running normally Docker Service or power supply , that PXC There is no time to judge who is the last node to exit , all PXC The nodes are closed in an instant , Which node is safe_to_boostrap The parameters are 0. It's also easy to solve this problem , That's picking node1, Change this parameter to 1, Then start up normally node1, Just start other nodes .

Four 、PXC The cluster has only one node , Close the container of this node , Can it be started next time ？

Of course you can , because PXC There is only one node in , Then this node must be started according to the master node , So when you start it , It will start MySQL service , Also created PXC colony . Even if the container is closed , Next time, it's still this step , There will be no startup failure . if PXC A cluster is composed of multiple nodes ,node1 It stopped , Other nodes are running normally . It's time to start node1 There will be a flash back ,node1 It just started for a few seconds and hung up . This is because node2 Wait for some nodes to be in the existing PXC Run in , At this time, you start node1, Create another one with the same name PXC colony , It will definitely lead to conflict . therefore node1 And then he stepped back .

In this case , The correct solution is , hold node1 Container delete . take it easy , Didn't let you delete v1 Data volume , So the data cannot be lost . Then create a node1, For the setting synchronized with a node in the startup parameters, choose any one that is running now PXC node , then Node1 You can start .