SecurityMigrationExecutor and SystemIndexMigrationExecutor should not be persistent tasks · elastic/elasticsearch#146662

(3 留言) (0 反應) (0 負責人)Java (76,700 star) (25,882 fork)batch import

:Distributed/DistributedTeam:Distributedgood first issue

描述

Both SecurityMigrationExecutor and SystemIndexMigrationExecutor are run-to-completion one-shot migration tasks rather than continuously-running services. They do not "persist" in any meaningful sense, instead they have some external retry mechanism, so it seems that they would better fit as just regular transport actions instead.

Refactoring those task to not be persistent tasks would also get us one step closer to removing the automaticReassignmentOnShutdown() persistent task executor flag (see https://github.com/elastic/elasticsearch/issues/138565) and making this feature default everywhere.

See: https://github.com/elastic/elasticsearch/pull/145753 for more details

貢獻者指南

技術棧: java
領域: backend
議題類型: refactor
難度: 4
預計時間: over 1 week
活動狀態: active
清晰度: mostly clear
前置要求: Understanding of Elasticsearch task execution frameworkJava development experience
新手友善度: 20
研究方向: Review the linked PR (https://github.com/elastic/elasticsearch/pull/145753) for detailed context on why SecurityMigrationExecutor and SystemIndexMigrationExecutor are currently persistent tasks and how to refactor them to regular transport actions. Examine the `automaticReassignmentOnShutdown()` flag mentioned in issue #138565 to understand the broader goal. The refactoring should ensure that these one shot migration tasks are no longer treated as continuously running services, simplifying the task execution model.