[IGNITE-21390] Inconsistent behavior of Compute APIs when target node does not exist

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0-beta1
• Fix Version/s: 3.0.0-beta2
• Component/s: compute, thin client
• Labels: ignite-3
• Assignee: Vadim Pakhnushev
• Reporter: Pavel Tupitsyn
• Created: 30/Jan/24 07:21
• Updated: 20/Feb/24 10:41
• Resolved: 20/Feb/24 10:41

Description

IgniteCompute.executeAsync accepts a set of nodes. If we pass a single node that does not exist in the cluster, the API behavior is confusing and inconsistent across embedded and thin client modes:

        var fakeNode = new ClusterNodeImpl("fakeId", "fakeName", new NetworkAddress("localhost", 12345));

        JobExecution<String> execution = ignite.compute().executeAsync(Set.of(fakeNode), units(), "job", null);

        execution.resultAsync().join();

• Client: throws "Specified node is not present in the cluster" exception with a generic error code (uses deprecated constructor)
• Embedded: actually tries to connect to the specified address and throws "io.netty.channel.AbstractChannel$AnnotatedConnectException: Connection refused: localhost/127.0.0.1:12345"

Ensure consistent behavior; probably add a dedicated error code.

Issue Links

• links to: GitHub Pull Request #3191