clustering questions

[SalmasCM <sa...@criticalmass.com>]

We are deploying a JR implementation which will consist of a JR cluster of
two independent web apps. One is an admin app and the other a consumer web
site. During testing I have been running both on one box. Now that we are
getting closer to production I want to verify that the observation api will
still work if the apps are deployed independently. I am assuming that the
notifications are handled the same way as the journal updates (through the
common db). Please let me know if I am understanding correctly how
clustering works.

Regards
-- 
View this message in context: http://www.nabble.com/clustering-questions-tp23127048p23127048.html
Sent from the Jackrabbit - Users mailing list archive at Nabble.com.

Re: clustering questions

[Alexander Klimetschek <ak...@day.com>]

On Sun, Apr 19, 2009 at 10:28 PM, SalmasCM <sa...@criticalmass.com> wrote:
>
> We are deploying a JR implementation which will consist of a JR cluster of
> two independent web apps. One is an admin app and the other a consumer web
> site. During testing I have been running both on one box. Now that we are
> getting closer to production I want to verify that the observation api will
> still work if the apps are deployed independently. I am assuming that the
> notifications are handled the same way as the journal updates (through the
> common db). Please let me know if I am understanding correctly how
> clustering works.

Yes, the journal is required to propagate the repository changes to
the other cluster nodes (not only for observation, but also for
cache-invalidation). If you use a db journal which is accessible by
all nodes it should work fine.

BTW, do you see anything missing from the description at
http://wiki.apache.org/jackrabbit/Clustering ?

Regards,
Alex

-- 
Alexander Klimetschek
alexander.klimetschek@day.com

Re: clustering questions

[Alexander Klimetschek <ak...@day.com>]

On Mon, Apr 20, 2009 at 9:28 PM, SalmasCM <sa...@criticalmass.com> wrote:
>
> Thanks for the verification. I do think that it would be nice to have some
> more details documented. These might include a section on the observation
> api and how it works with RMI and with clustering.
> Also, I found out the hard way that when working with multiple workspaces I
> should do a clone and then subsequent merges. it would be nice to have a
> document that shows some tips and tricks on how to use multiple workspaces.
> One issue we have is that we used the XML import feature to import large
> amounts of legacy data. Of course these were all imported as Strings. When
> we tried to set node type properties we found that there was no way to unset
> them or change them. Is it possible for us to
>
> 1. Save all our data in the db except these and persist these on the
> filesystem so that we could change them if needed?
> 2. We can always skip the xml load utility and import the data using the JR
> api using booleans and doubles and dates as required. Would this be
> advisable? I think setting node type properties would be a better thing to
> do. Some discussion or explanation of these issues would be great as well.

Changing the primary node type of a node will be possible with
JSR-283, but for now you can only assign it while creating a node. But
you have two other options to cope with changing properties: a) mixins
and b) residual properties aka "*" wildcard props. See the JCR spec
for more info [1] [2].

[1] http://jcp.org/en/jsr/detail?id=170
[2] http://www.day.com/specs/jcr/1.0/

Regards,
Alex

-- 
Alexander Klimetschek
alexander.klimetschek@day.com

Re: clustering questions

[SalmasCM <sa...@criticalmass.com>]

Thanks for the verification. I do think that it would be nice to have some
more details documented. These might include a section on the observation
api and how it works with RMI and with clustering. 
Also, I found out the hard way that when working with multiple workspaces I
should do a clone and then subsequent merges. it would be nice to have a
document that shows some tips and tricks on how to use multiple workspaces.
One issue we have is that we used the XML import feature to import large
amounts of legacy data. Of course these were all imported as Strings. When
we tried to set node type properties we found that there was no way to unset
them or change them. Is it possible for us to

1. Save all our data in the db except these and persist these on the
filesystem so that we could change them if needed?
2. We can always skip the xml load utility and import the data using the JR
api using booleans and doubles and dates as required. Would this be
advisable? I think setting node type properties would be a better thing to
do. Some discussion or explanation of these issues would be great as well.

Thanks


SalmasCM wrote:
> 
> We are deploying a JR implementation which will consist of a JR cluster of
> two independent web apps. One is an admin app and the other a consumer web
> site. During testing I have been running both on one box. Now that we are
> getting closer to production I want to verify that the observation api
> will still work if the apps are deployed independently. I am assuming that
> the notifications are handled the same way as the journal updates (through
> the common db). Please let me know if I am understanding correctly how
> clustering works.
> 
> Regards
> 

-- 
View this message in context: http://www.nabble.com/clustering-questions-tp23127048p23143296.html
Sent from the Jackrabbit - Users mailing list archive at Nabble.com.