[IGNITE-22725] (flaky) aimem: node doesn't recover after data clean and goes into state DEGRADED

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.0
• Component/s: persistence
• Labels: ignite-3

Environment

3 distributed nodes (with 1 CMG), storage AIMEM

• Ignite Flags: Docs Required, Release Notes Required
• Assignee: Unassigned
• Reporter: Igor
• Created: 12/Jul/24 16:08
• Updated: 04/Sep/24 16:22

Description

This issue sometimes is replaced by https://issues.apache.org/jira/browse/IGNITE-22724

Steps to reproduce:

1. Create 3 nodes cluster with 1 CMG node (node_0 - CMG,node_1,node_2).
2. Create zone with replication equals to amount of nodes (3).
3. Create 10 tables inside the zone.
4. Insert 100 rows in every table.
5. Await all tables*partitions*nodes local state is "HEALTHY"
6. Await all tables*partitions*nodes global state is "AVAILABLE"
7. Kill non CMG node with kill -9. (kill node_1)
8. Clean data in work directory of killed node (node_1).
9. Start killed node.
10. Using REST API await physical topology has 3 alive nodes.
11. Using REST API await logical topology has 3 alive nodes.
12. Await all tables*partitions*nodes local state is "HEALTHY" (by connecting to REST of node_2).
13. Await all tables*partitions*nodes local state is "AVAILABLE" (by connecting to REST of node_2).

Expected:
All partitions become "AVAILABLE".

Actual:
Partitions in states: "AVAILABLE" and "DEGRADED".server_logs.zip

Comments:

This test works fine with aipersist storage. Similar to https://issues.apache.org/jira/browse/IGNITE-22517 , but the scenario a bit different.

Issue Links

• is caused by: IGNITE-22757 Excessive memory usage in schema-related code in SQL