[jira] [Resolved] (MESOS-699) Slaves with RO filesystem shouldn't launch tasks requesting disk

["Vinod Kone (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/MESOS-699?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Vinod Kone resolved MESOS-699.
------------------------------
    Resolution: Fixed

Slave checkpointing is enabled by default now. If the work directory has RO file system the slave itself will flap before it can offer resources and run tasks.

> Slaves with RO filesystem shouldn't launch tasks requesting disk
> ----------------------------------------------------------------
>
>                 Key: MESOS-699
>                 URL: https://issues.apache.org/jira/browse/MESOS-699
>             Project: Mesos
>          Issue Type: Bug
>            Reporter: Joe Smith
>            Priority: Minor
>
> At Twitter, we had a few slave with errors in its hard drive, yet it still sent resource offers. Unfortunately, when tasks attempted to start up, the executor failed when writing its sandbox, as the disk was RO:
> {code}
> IOError: [Errno 30] Read-only file system
> {code}
> It'd be awesome if the slave could detect that and not launch tasks that request disk.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)