[IGNITE-10513] Java client stucks when connects to server with slow disk

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.6
• Fix Version/s: 2.8
• Component/s: general
• Labels: discovery
• Assignee: Ilya Kasnacheev
• Reporter: Dmitry Lazurkin
• Created: 03/Dec/18 16:43
• Updated: 11/Dec/18 10:36
• Resolved: 11/Dec/18 10:24

Description

For emulating slow disk add sleep to partitions cycle in GridCacheDatabaseSharedManager#restorePartitionStates:

//...
            for (int i = 0; i < grp.affinity().partitions(); i++) {
                 try {
                     log.error("Wait");
                     Thread.sleep(10000);
                 } catch (InterruptedException e) {
                     e.printStackTrace();
                 }
//...

My server has 1024 partitions.

Steps to reproduce:

• Start server
• Start client
• On client wait message "Join cluster while cluster state transition is in progress, waiting when transition finish."
• Kill server
• On client wait repeatable java.net.ConnectException: Connection refused (Connection refused)
• Start server (I have 100% chance to reproduce issue on my computer)

Issue Links

• is related to: IGNITE-4641 Refresh client attributes during reconnect
• links to: GitHub Pull Request #5557
• links to: Ignite User Mailing List
• links to: TC run on Master
• links to: TC Run with Zk change