Making sure that the cluster is able to tolerate a node going down is crucial because we can confirm that no downtime occurs if a node is lost.
This can be done by forcibly shutting down one of the nodes while the others continue to serve data. To function as a synthetic workload, we can use FIO to perform a continuous test while one of the nodes is being shut down.
In the following screenshot, we can see that the gfs2 node was not present, but the FIO test continued serving data as expected: