[ https://issues.apache.org/jira/browse/IGNITE-19971?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17746256#comment-17746256 ]
Vladislav Pyatkov commented on IGNITE-19971: -------------------------------------------- LGTM > Flaky ItNodeTest#testLeaseReadAfterSegmentation > ----------------------------------------------- > > Key: IGNITE-19971 > URL: https://issues.apache.org/jira/browse/IGNITE-19971 > Project: Ignite > Issue Type: Bug > Reporter: Mirza Aliev > Assignee: Mirza Aliev > Priority: Major > Labels: ignite-3 > Time Spent: 10m > Remaining Estimate: 0h > > ItNodeTest#testLeaseReadAfterSegmentation is flaky > {code:java} > java.lang.NullPointerException: Cannot invoke > "org.apache.ignite.raft.jraft.Node.getNodeId()" because the return value of > "org.apache.ignite.raft.jraft.core.TestCluster.getLeader()" is null > at > org.apache.ignite.raft.jraft.core.ItNodeTest.lambda$testLeaseReadAfterSegmentation$43(ItNodeTest.java:3652) > {code} > The problem was in the race in > {code:java} > assertTrue(waitForCondition(() -> cluster.getLeader() != null && > !leader.getNodeId().equals(cluster.getLeader().getNodeId()), > 10_000)); > {code} > so leader was stepped done in the middle of this assertion, so the first part > of this expression {{cluster.getLeader() != null}} was true. > The fix is to rewrite condition to defence against NPE. > After 300 local runs there is no failures, but 3 failures in 100 runs were > before fix -- This message was sent by Atlassian Jira (v8.20.10#820010)