[HDFS-11633] FSImage failover disables all erasure coding policies

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 3.0.0-alpha4
• Fix Version/s: 3.0.0-alpha4
• Component/s: erasure-coding, namenode
• Labels: hdfs-ec-3.0-must-do
• Target Version/s: 3.0.0-alpha4
• Assignee: Wei-Chiu Chuang
• Reporter: Wei-Chiu Chuang
• Created: 07/Apr/17 07:18
• Updated: 21/May/20 09:21
• Resolved: 07/Apr/17 22:47

Description

If NameNode fails to load the fsimage in the first NameNode metadata directory, it accidentally clears all enabled erasure coding policies in ErasureCodingPolicyManager.

Even if the NameNode configures multiple fsimage metadata directory and it successfully loads the second one, enabled erasure coding policies are not restored.

In the current implementation, we do not have a ErasureCodingPolicyManager section in fsimage, so a fsimage reload does not reload ECPM.

The easiest fix before ECPM section is implemented, is don't clear ECPM when FSNamesystem is cleared.