[
https://issues.apache.org/jira/browse/MESOS-8446?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16328204#comment-16328204
]
Qian Zhang commented on MESOS-8446:
-----------------------------------
commit 2c5da1b668de91e33831caafb18a3b4d71b26c69
Author: Qian Zhang
Date: Mon Jan 15 16:40:00 2018 +0800
Detached `virtualLatestPath` when recovering the executor.
Previously we miss to detach `/frameworks/FID/executors/EID/runs/latest`
when we find the latest run of the executor was completed in the method
`Framework::recoverExecutor`, that is a leak.
Review: https://reviews.apache.org/r/65167
> Agent miss to detach `virtualLatestPath` for the executor's sandbox during
> recovery
> -----------------------------------------------------------------------------------
>
> Key: MESOS-8446
> URL: https://issues.apache.org/jira/browse/MESOS-8446
> Project: Mesos
> Issue Type: Bug
> Components: agent
> Reporter: Qian Zhang
> Assignee: Qian Zhang
> Priority: Major
>
> In {{Framework::recoverExecutor()}}, we attach {{executor->directory}} to 3
> virtual paths:
> (1) /agent_workdir/frameworks/FID/executors/EID/runs/CID
> (2) /agent_workdir/frameworks/FID/executors/EID/runs/latest
> (3) /frameworks/FID/executors/EID/runs/latest
> But in this method, when we find the executor completes, we only do detach
> for (1) and (2) but not (3). We should do detach for (3) too as what we do in
> {{Slave::removeExecutor}}, otherwise, it will be a leak.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)