curator-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CURATOR-84) More flexibility for InterProcessMutex extensions
Date Mon, 18 Aug 2014 22:57:19 GMT

    [ https://issues.apache.org/jira/browse/CURATOR-84?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14101463#comment-14101463
] 

ASF GitHub Bot commented on CURATOR-84:
---------------------------------------

Github user karkum commented on the pull request:

    https://github.com/apache/curator/pull/38#issuecomment-52567036
  
    Hmm, let me make sure I understand:
    
    * In ```InterProcessMutex```, we expose the actual lockPath being used by the lock (which
could be either ephemeral or persistent).
    * This would allow any user of ```InterProcessMutex``` to take care of clean up/notification
steps based on whether the lock of that path actually exists or not.
    
    I think this makes sense, but I'm not sure if it is ideal to expose the actual lock path
used by this recipe. What if a client is using the default implementation (with ephemeral
nodes), then getting the path would be pretty useless, since that node would be deleted once
connection is lost right? 


> More flexibility for InterProcessMutex extensions
> -------------------------------------------------
>
>                 Key: CURATOR-84
>                 URL: https://issues.apache.org/jira/browse/CURATOR-84
>             Project: Apache Curator
>          Issue Type: Wish
>          Components: Recipes
>    Affects Versions: 2.3.0
>            Reporter: Jozef Vilcek
>         Attachments: CURATOR-84.patch
>
>
> I have a need for a durable InterProcessMutex. Main reason for this are processes with
critical sections, where I can not afford to loose a lock due to session expiration. In such
case, others might acquire a lock and kick in while the previous process is still running
but e.g. experiencing connection issues. To kill this temporally detached process in favor
of others would be too costly.
> To achieve such behavior, I need lock nodes to be created in PERSISTENT mode. This is
not possible to do easily with currently implementation of locks due to few internal scoped
classes and methods. I would like to change this.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message