helix-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Zhen Zhang (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HELIX-254) Helix controller should disable participants with mismatched session-id
Date Thu, 26 Sep 2013 22:53:02 GMT

     [ https://issues.apache.org/jira/browse/HELIX-254?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Zhen Zhang updated HELIX-254:
-----------------------------

    Description: 
Participants may have a mismatched session-id in live-instance (session-id filed in live-instance
znode is different from the ephemeral owner id in live-instance znode's meta data) due to
race conditions in zk session expiry (e.g. two consecutive long gc). this causes problems
when controller sending transition messages to participant and participant just removes the
messages due to mismatched session-id. this will lead to infinite loop and flooding zk server.

While we have a ticket for fixing the mismatched session-id issue, it will be also safe to
have a check on helix controller for liveinstance changes. if a liveinstance with mismatched
session-id is detected, helix controller should disable the participant.

  was:
Participants may have a mismatched session-id in live-instance (session-id filed in live-instance
znode is different from the ephemeral owner id in live-instance znode's meta data). this causes
problems when controller sending transition messages to participant and participant just removes
the messages due to mismatched session-id. this will lead to infinite loop and flood zk server.

While we have a ticket for fixing the mismatched session-id issue, it will be also safe to
have a check on helix controller for liveinstance changes. if a liveinstance with mismatched
session-id is detected, helix controller should disable the participant.

    
> Helix controller should disable participants with mismatched session-id
> -----------------------------------------------------------------------
>
>                 Key: HELIX-254
>                 URL: https://issues.apache.org/jira/browse/HELIX-254
>             Project: Apache Helix
>          Issue Type: Bug
>            Reporter: Zhen Zhang
>            Assignee: Zhen Zhang
>
> Participants may have a mismatched session-id in live-instance (session-id filed in live-instance
znode is different from the ephemeral owner id in live-instance znode's meta data) due to
race conditions in zk session expiry (e.g. two consecutive long gc). this causes problems
when controller sending transition messages to participant and participant just removes the
messages due to mismatched session-id. this will lead to infinite loop and flooding zk server.
> While we have a ticket for fixing the mismatched session-id issue, it will be also safe
to have a check on helix controller for liveinstance changes. if a liveinstance with mismatched
session-id is detected, helix controller should disable the participant.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message