[CASSANDRA-14793] Improve system table handling when losing a disk when using JBOD

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 4.0-rc1, 4.0
• Component/s: Legacy/Core
• Severity: Normal
• Complexity: Normal
• Since Version: NA

Source Control Link

https://github.com/apache/cassandra/commit/f69b11eee9605add3a006de46eedb773a984d90b

Test and Documentation Plan

• Run of JUnit tests with separate directory for local system keyspaces
• New unit tests for some of the new code.

• Assignee: Benjamin Lerer
• Reporter: Marcus Eriksson

Authors

Benjamin Lerer

Reviewers

Andres de la Peña, Marcus Eriksson

• Created: 26/Sep/18 15:32
• Updated: 16/Mar/22 14:01
• Resolved: 04/Feb/21 14:14

Description

We should improve the way we handle disk failures when losing a disk in a JBOD setup

 One way could be to pin the system tables to a special data directory.

Issue Links

• links to: PR trunk