multiclient_test/

Multi-client Robustness Test

The Kopia multi-client robustness test is intended to verify the correct restoration of application persistent data over time. The test design is similar to the standard Kopia robustness test, reusing the main test engine and most of its components. However, this test suite uses Kopia's client-server mechanisms to run multiple clients concurrently.

Components

The multi-client Kopia robustness test has the following components:

• MultiClientFileWriter
  This is an implementation of the robustness.FileWriter interface that creates and manages client FileWriter's. In addition, the MultiClientFileWriter delegates FileWriter function calls to a client's specific FileWriter.

• MultiClientSnapshotter
  This is an implementation of the Snapshotter interface that creates and manages a single Server and one ClientSnapshotter per client. In addition, the MultiClientSnapshotter delegates Snapshotter function calls to a client's specific ClientSnapshotter.

• Server
  This is the component responsible for setting up and managing the Kopia server. It has the following functionality:

  • Set up or connect to a Kopia repository and create a server for clients to connect to
  • Add and remove permissions for new clients
  • Run commands directly on the Kopia repo, such as garbage collection

• ClientSnapshotter
  This is a Snapshotter with the added functionality of being able to connect to and disconnect from a server.

The diagram below shows the standard Kopia robustness test framework on the left and the multi-client framework on the right. It illustrates how the components described above fit into the engine and delegate actions to client instantiations of the FileWriter and ClientSnapshotter components.

Creating Multi-client Tests

The file multiclient_test.go contains the existing tests, which can be used as an example. The snippet below shows how to use the multi-client framework to create a test with four concurrent clients.

func TestExample(t *testing.T) {
	numClients := 4

	// Define per-client test actions
	f := func(ctx context.Context, t *testing.T) {
		t.Helper()

		_, err := eng.ExecAction(ctx, /* engine-action-key */, /* opts */)
		// ... other actions ...
	}

	// Create or obtain root context
	ctx := testlogging.Context(t)

	// Run test actions for each client concurrently
	th.RunN(ctx, t, numClients, f)
}