[kafka] branch 2.8 updated: MINOR: Assume unclean shutdown for metadata log (#10363)

[jg...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

jgus pushed a commit to branch 2.8
in repository https://gitbox.apache.org/repos/asf/kafka.git


The following commit(s) were added to refs/heads/2.8 by this push:
     new 3b6ab58  MINOR: Assume unclean shutdown for metadata log (#10363)
3b6ab58 is described below

commit 3b6ab58c8df823269d6bdc73820618d45ac147e5
Author: José Armando García Sancio <js...@users.noreply.github.com>
AuthorDate: Fri Mar 19 15:09:59 2021 -0700

    MINOR: Assume unclean shutdown for metadata log (#10363)
    
    Currently the metadata log assumes successful shutdown and skips recovery. For now, we prefer to err on the safe side and assume instead that the replica was shutdown uncleanly so that we can force full recovery. This is justified in the short term because:
    
    1. Snapshots are not fully implemented for the metadata log
    2. The replicas (controllers and brokers) need to read the entire metadata log to load it into memory.
    
    In other words, we need to read through the metadata log once on startup anyway. A long-term fix will be provided in https://issues.apache.org/jira/browse/KAFKA-12504.
    
    Reviewers: Jason Gustafson <ja...@confluent.io>
---
 core/src/main/scala/kafka/raft/KafkaMetadataLog.scala | 1 +
 1 file changed, 1 insertion(+)

diff --git a/core/src/main/scala/kafka/raft/KafkaMetadataLog.scala b/core/src/main/scala/kafka/raft/KafkaMetadataLog.scala
index 64ae636..783bd4c 100644
--- a/core/src/main/scala/kafka/raft/KafkaMetadataLog.scala
+++ b/core/src/main/scala/kafka/raft/KafkaMetadataLog.scala
@@ -315,6 +315,7 @@ object KafkaMetadataLog {
       maxProducerIdExpirationMs = Int.MaxValue,
       producerIdExpirationCheckIntervalMs = Int.MaxValue,
       logDirFailureChannel = new LogDirFailureChannel(5),
+      lastShutdownClean = false,
       keepPartitionMetadataFile = false
     )