Jepsen对Orkes Conductor工作流编排系统进行了分布式系统安全性测试，在5.3.2、5.4.0及5.5.0-rc等多个版本中发现了9个主要问题1。测试在包含单个Postgres节点、Redis或Redis Cluster、多个Conductor服务器和worker节点的10台Debian机器集群环境中进行1。
发现的缺陷包括任务并发执行、工作流执行停滞、连接失败返回但工作流实际执行、工作流被永久放弃、工作流丢失以及标记为完成但从未执行等问题1。其中，在健康集群中任务并发执行频率约每6小时一次，减少租赁时间到1秒后每几分钟发生一次1。当使用Postgres数据库加Redis队列配置时，工作流可在任务执行前停滞长达默认响应超时时间1小时1。
Orkes已在版本5.5.6中修复了其中部分问题1，而问题#3和#6仍在调查中1。针对涉及工作流被永久放弃、丢失以及标记为完成但从未执行的严重缺陷，建议用户升级到版本5.5.61。
Jepsen, a distributed systems testing framework, has uncovered nine significant defects in Orkes Conductor workflow orchestration system across versions 5.3.2, 5.4.0, and 5.5.0-rc during comprehensive security testing 1. The vulnerabilities span multiple failure modes, including concurrent task execution in healthy clusters, workflow execution stalls, null workflow ID returns, SQL syntax errors, duplicate key errors, and scenarios where connections fail despite workflows actually executing 1.
Among the discovered issues, three were classified as severe security problems: workflows being permanently abandoned, workflows disappearing from client visibility, and workflows marked as complete despite never having executed 1. Testing was conducted on a 10-node Debian cluster environment incorporating a single PostgreSQL node, Redis or Redis Cluster instances, multiple Conductor servers, and worker nodes using the Conductor Java SDK version 5.0.1 1. The issues manifested at varying frequencies—for instance, concurrent task execution occurred approximately every six hours under normal conditions, but increased to several times per minute when lease time was reduced to one second 1. In Postgres-plus-Redis configurations, workflows exhibited stalls before task execution for periods lasting up to the default one-hour response timeout 1.
Orkes has addressed four of these problems in version 5.5.6 1. Two additional issues remain under investigation 1. The company recommends users upgrade to version 5.5.6 to resolve the three critical security vulnerabilities 1.
评论
还没有评论，欢迎留下第一条。