You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@mesos.apache.org by "Chi Zhang (JIRA)" <ji...@apache.org> on 2015/02/18 20:58:12 UTC

[jira] [Commented] (MESOS-2370) Better management of ingress qdisc on eth0 and lo

    [ https://issues.apache.org/jira/browse/MESOS-2370?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14326465#comment-14326465 ] 

Chi Zhang commented on MESOS-2370:
----------------------------------

cc [~jieyu]

> Better management of ingress qdisc on eth0 and lo
> -------------------------------------------------
>
>                 Key: MESOS-2370
>                 URL: https://issues.apache.org/jira/browse/MESOS-2370
>             Project: Mesos
>          Issue Type: Bug
>            Reporter: Chi Zhang
>            Priority: Minor
>
> PortMappingIsolator currently does not keep track of whether it is the one that adds the ingress qdiscs and never cleans them up.
> The goal is to not interleave with other administrators on the system but try to clean up after ourselves reasonably. A few steps we can do to make it better:
> maintain a bool and initialize to be false. True means we think someone else on the system is using the ingress qdiscs.
> in create, if there are already ingress qdiscs, set the bool to true; otherwise, create them.
> in cleanup, after cleaning up after our last container, if there are other filters, set the bool to true and continue; however, if there are no other filters and the bool is false, clean up the ingress qdiscs. This way after a proper shutdown, the system is clean.
> in isolate, we repeat the same behavior as in create.
> Overall, this shouldn't do worse than the current behavior but try to safely clean up after ourselves.
>  



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