[jira] [Created] (IGNITE-18139) Fix NPE produced by a call from InternalTableImpl#enlistWithRetry()

["Roman Puchkovskiy (Jira)" <ji...@apache.org>]

Roman Puchkovskiy created IGNITE-18139:
------------------------------------------

             Summary: Fix NPE produced by a call from InternalTableImpl#enlistWithRetry()
                 Key: IGNITE-18139
                 URL: https://issues.apache.org/jira/browse/IGNITE-18139
             Project: Ignite
          Issue Type: Bug
            Reporter: Roman Puchkovskiy
            Assignee: Alexander Lapin
             Fix For: 3.0.0-beta2


I'm doing the following:
 # Start a cluster of 3 nodes
 # Create a table with 3 replicas and 1 partition
 # Stop one node
 # Make a snapshot on the leader of the partition (twice in a row) - I'm not sure this step is required to reproduce the issue
 # Start the stopped node

Then in this code
private <R> CompletableFuture<R> enlistWithRetry(
        InternalTransaction tx,
        int partId,
        BiFunction<TablePartitionId, Long, ReplicaRequest> requestFunction,
        int attempts
) {
    CompletableFuture<R> result = new CompletableFuture<>();

    enlist(partId, tx).<R>thenCompose(
                    primaryReplicaAndTerm -> {
                        try {
                            return replicaSvc.invoke(
                                    primaryReplicaAndTerm.get1(),
                                    requestFunction.apply((TablePartitionId) tx.commitPartition(), primaryReplicaAndTerm.get2())
                            );
{{primaryReplicaAndTerm}} turns out to contain null as first element, 1 as second element, which causes an NPE further. Probably, {{clusterNodeResolver}} returned null.

This does not reproduce often, but I saw this a few times.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)