It's nothing else today , I just want to make trouble ETCD What will happen if the cluster fails , Have fun .
0. The environment is , Three independent ubuntu machine , Build a three node etcd colony .
1. Let's take a look at a set of normal operations .
2. Kill a node directly , But it's not the master node .
At this point, it is executed on two living machines etcdctl member list This is the result
Then the three stations will execute the operation separately etcdctl endpoint health, The two live are health, The dead are unhealth.
And then on one of the two living machines put A value , The other one is OK get Arrived .
3. I'm trying to start the third machine that died , then get In the next step put Try it .
No problem. You can synchronize it .
4. The next step is to kill two
dying , It's useless , One is out of operation
Take a look at the last one etcd The window that started .
The following information scrolls over and over again
Save one at will, and then you can operate it , This is not a screenshot .
In fact, the backstage is always looking for the third dead machine
After starting the third channel , No rolling in a moment , It's gone . It's all connected
5. Last but not least , namely etcdctl member list
This command doesn't mean to get online node information , This place is easy to misunderstand , For example, I've turned it off now 108 This node （ Shut down completely , It's not a simple exit etcd）
OK That's all I just want to see what happens , Operation, operation and play . Finishing the next article ratf Algorithm correlation .