Igor created IGNITE-21663: ----------------------------- Summary: Cluster load balancing when 1 node is killed doesn't work Key: IGNITE-21663 URL: https://issues.apache.org/jira/browse/IGNITE-21663 Project: Ignite Issue Type: Improvement Components: persistence, sql Affects Versions: 3.0.0-beta1 Reporter: Igor Fix For: 3.0.0-beta1
*Steps to reproduce:* # Start cluster with 2 nodes running locally. # Make connection like this: {code:java} try (IgniteClient igniteClient = IgniteClient.builder().retryPolicy(new RetryLimitPolicy()).addresses(thinClientEndpoints.toArray(new String[]{"localhost:10800","localhost:10801"})).build()) { try (Session session = igniteClient.sql().createSession()) { //code here } } {code} 3. Create table with replication 2 4. Execute insert 1 row and select from the table. 5. Kill first node (in list of connection) 6. Execute select from the table. *Expected:* Cluster works with one node. *Actual:* The exception, the selec is not executed. *Comments:* The Java client makes request to the working node, but the working node tries to connect to the killed one and gets connection exception. -- This message was sent by Atlassian Jira (v8.20.10#820010)