openjpa-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jaeger, Michael" <michael.c.jae...@siemens.com>
Subject AW: dual node URLs (of SolidDB)
Date Thu, 04 Dec 2008 09:20:41 GMT
Hi,

regarding the WARN message: there is a WARN message (or at least it was in one of the previous
versions of OpenJPA) if there is a Dictionary in use that does not belog to the openjpa-package
for dictionaries. Actually, we understand the sense of this message, but what happens is the
following: We have put openjpa into the OSGi environemnt, so other developers can use this
one when developing their components. And we have put an additional dictionary for solid in.

Once, an developer experiences a problem with using the O/R-mapping, sHe complains: "... and
I have seen a WARN message saying that you guys have put in something unsupported by openjpa,
any chance that the actual error has something to do with this?" When we assure that the error
is not because of our dictionary, we get the question back, why there is a WARN message then.
So we have psychological problem here with messages at ll WARN... solution is to put the dict
into your package, however, we are not happy with this.

regarding the submission of the solid dictionary I have forwarded the inquiry and will come
back on this.

kind regards, Michael


-----Urspr√ľngliche Nachricht-----
Von: Kevin Sutter [mailto:kwsutter@gmail.com] 
Gesendet: Mittwoch, 3. Dezember 2008 16:09
An: users@openjpa.apache.org
Betreff: Re: dual node URLs (of SolidDB)

Michael,
This is cool.  I have had a hidden desire to pursue support of SolidDB with
OpenJPA, but I just haven't found the time.  This is very good news that
your initial experiences are positive.

What dictionary are using with SolidDB?  It sounds like you might have
created a custom dictionary from your note, but you must have started with
one as a base.  Have you tried running the OpenJPA regression bucket against
your SolidDB dictionary?  Minus the dual url, of course.  I'm just curious
how close we are to having a dictionary for SolidDB that could be used for
our regression runs.  I would assume that the use of SolidDB would speed up
this processing to some extent.

Would you be interested in contributing your custom SolidDB dictionary?  I
hate to beg.  :-)  But, like I said, this has been an interest of mine for
quite some time...

We would need to take a closer look at the proposed patch for
parseProperties().  Off the top of my head, I wonder whether the stop
parsing at the comma was due to a problem reported in the past, or if it was
just thought to be the right thing to do.  We would need to determine
whether this comma-delimited url would be an option, or just allow anything
in the url.  Nothing major, but a few things to check on.

I guess I am not familiar with the message you are referring to with regards
to the wrong package being used.  Can you elaborate?

I had opened a JIRA Issue some time ago for SolidDB (
https://issues.apache.org/jira/browse/OPENJPA-735).  We could use this Issue
for any proposed patches, dictionaries, and discussion.

Thanks!
Kevin



On Wed, Dec 3, 2008 at 6:41 AM, Jaeger, Michael <
michael.c.jaeger@siemens.com> wrote:

> Hi all,
>
> we know that soliddb is not among the supported databases of OpenJPA.
> Thus we would like place our experience with soliddb + openjpa to
> discussion (if support for SolidDB is anticipated).
>
> The main reason why we use soliddb is its hot-standby-feature. For
> setting up hot standby, just pass the JDBC driver a dual node URL.
> Solidtech has decided to separate the adresses of two nodes by a comma
> where normally a single host is named. looks like this:
>
>        Url=jdbc:solid://comp1:1315,comp2:1315/myusr/mypw
>
> so the point with OpenJPA here is that properties are separated by ","
> as well, so the rest of the URL is cut away then:
>
>        Url=jdbc:solid://comp1:1315
>        comp2:1315/usr/pw
>        DriverClassName=solid.jdbc.SolidDriver
>
> We have patched this (Configurations.parseProperties()), simply inject
> the complete URL at some later point again.
>
> Clearly, the support for malformed URLs can be put into question. thus,
> on first hand, take our e-mail as feedback where one major point would
> come up when using SolidDB. We have had also some other issue which we
> fixed with a custom dictionary. In summary our experience with using
> openjpa + soliddb is very positive apart from the fundamental issue with
> the dual node URL.
>
> - Michael
>
> PS. the warning message when using a dictionary not belonging to the
> right package is also not very nice and here on our site has disturbed
> the developers. anyone experience with this one?
>

Mime
View raw message