jackrabbit-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bart van der Schans (Commented) (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (JCR-2968) Add an option to read the Clustering Journal from a different source than the rest of the clustering info
Date Wed, 12 Oct 2011 08:45:11 GMT

    [ https://issues.apache.org/jira/browse/JCR-2968?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13125668#comment-13125668
] 

Bart van der Schans commented on JCR-2968:
------------------------------------------

Hi Jukka, Christian,

I agree about the code duplication. So I'm merging Christian's and mine solution to the "run
jr on a db slave" issue. 

The current "logic" is:
1. With the "normal" (old) cluster/journal configuration everything works as it always did
2. If the readOnly flag (from the patch from Christian) is set to true, the FileRevision is
used; unless
3. An explicit "InstanceRevision" (database connection) is configured

Number 2 is what I proposed and number 3 is the proposal of Christian.

To achieve this I have:
- split out the DatabaseRevision part from the journal code
- split out the DatabaseConfig from the journal code in o.a.j.core.util.db so it can be reused
by the DatabaseRevision (and maybe other parts of the code as well)
- added the readOnly property
- use a FileRevision when the readOnly property is set

(see https://github.com/schans/jackrabbit/commits/JCR-2968 , still work in progress)

For the last bit I would like your opinions on how to specify the database instance revision
connection. I'm thinking about adding an "InstanceRevision" directly as child of the "Journal"
config element in the repository.xml, but it can also be a child of the Cluster element itself.
The InstanceRevision element will the same db parameters as the Journal element.

Wdyt?


                
> Add an option to read the Clustering Journal from a different source than the rest of
the clustering info
> ---------------------------------------------------------------------------------------------------------
>
>                 Key: JCR-2968
>                 URL: https://issues.apache.org/jira/browse/JCR-2968
>             Project: Jackrabbit Content Repository
>          Issue Type: New Feature
>          Components: clustering, jackrabbit-core
>    Affects Versions: 2.2.9
>            Reporter: Christian Stocker
>            Assignee: Jukka Zitting
>              Labels: patch
>         Attachments: 0001-JCR-2968-Add-an-option-to-read-the-Clustering-Journa.patch,
database-slave-local-revision-on-file.diff, patch_commit_2eed44310e71.patch
>
>
> This patch adds the possibility to read (but not write) the Cluster JOURNAL from a different
source than the rest of the cluster information. This makes it possible to setup a master/slave
DB setup, where everything cluster related is read from the slave, but writes to the master.
It reads the actual data also from the slave and assumes that this jackrabbit instance never
does any writes (except for updating the cluster index position in the DB). We have to read
the Cluster Journal from the slave to guarantee a consistent state
> More info why and how is here
> http://blog.liip.ch/archive/2011/05/04/how-to-make-jackrabbit-globally-distributable-fail-safe-and-scalable-in-one-go.html
> I didn't write any tests yet, if you can point me, where I should add them, I'll gladly
do them.
> Would be great, if we could integrate that in any of the future Jackrabbit releases.
> It's of course fully backwards compatible, nothing changes, if you don't sepcify 
>  <param name="dataSourceNameJournalRead">
> in repository.xml

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message