Using two RMI configurations for a single server
Hello, I am running two instances of Orion on my backend segment. Both servers run their own EJB applications. They are not clustered, shared or anything like that and they accept RMI connections to separate ports. Then I have one instance of Orion running on front-end. This instance is deployed with two web applications. The setup I'am trying to accomblish is therefore following: front-end back-end +--+ | EJB 1 | +---+--+--+ |WebApp1|---/ | | |WebApp2|---+--+ +---+ \ | EJB 2 | --+--+ The problem is that I haven't found any way to specify RMI configuration per-application. As I understood it, it can only be done per-server. A path to rmi.xml is specified in server.xml. JNDI can be easily setup per application in many ways but is there way to do the same for RMI configuration? If not, I think I have to split up the frontend server instance too? Thanx! -- Joni [EMAIL PROTECTED] http://www.shiftcontrol.com
ORMI problem
Hi all, I've noticed a strange bug in our latest production environment relating to the behaviour of Orion's RMI protocol (ORMI). I'm running Orion's web containers in different network segment than Orion's ejb containers. So, the RMI traffic is going through a firewall which is between these segments. When there has been lots of idle time, the front-end web container loses the RMI connection and hangs on it without ever trying to make a new connection. Has any of you noticed similar behaviour or do you have any guesses what might be going on? Otherwise I have to set up a stupid ping application which periodically makes HTTP requests and keeps the connection alive ;) -- Joni [EMAIL PROTECTED]
Re: Some questions on Orion
Kevin Duffey wrote: I will get my try at LDAP soon and I recall seeing some posts here on LDAP, so I would imagine it works with LDAP..of what company I have no idea. I've been using OpenLDAP (http://www.openldap.org) with Sun's LDAP provider in Linux. So far I haven't had any problems. The LDAP operations I've done are very basic though (listing, adding and deleting entries). -- Joni [EMAIL PROTECTED]
Re: SV: LDAP and EJBs
Thanks Patrik!
There's one issue which still puzzles me. Do I have to write my own
rollback code for LDAP operations when a transaction rolls back? I
browsed the JNDI API and SPI and I couldn't figure out if JNDI somehow
supports transactions (like JDBC does with begin(), rollback() and
commit()).
If my code is following for instance:
UpdatingFacadeEJB implements SessionBean {
public void doTheUpdate() {
ldapDelegate.writeLdapEntry();
rdbmsDelegate.writeRdbmsRow();
}
}
Let's say that ldapDelegate.writeLdapEntry() succeeds but
rdbmsDelegate.writeRdbmsRow() fails by throwing an EJBException. In this
case can the container's transaction manager somehow rollback the LDAP
operations done earlier in ldapDelegate.writeLdapEntry() method?
--
Joni
[EMAIL PROTECTED]
Patrik Andersson wrote:
You would have one sessionbean being business delegate in front of the
EntityBean manipulating the RDBMS and another managing the LDAP
directory service. Then you would have a Facade in front of both these
sessionbeans with one method that does the whole trick with
transactional attributes set up to require a transaction before
entering the method.
interface UpdatingFacade : SessionBean {
void doTheUpdate() throws RemoteException;
}
interface LdapDelegate : SessionBean {
void writeLdapEntry() throws RemoteException;
}
interface RdbmsDelegate : SessionBean or EntityBean {
void writeRdbmsRow() throws RemoteException;
}
Then you want an error in the writeLdapEntry() aswell as the
writeRdbmsRow() to throw an EJBException breaking the transaction
begun by doTheUpdate() method. Both writeXxx methods must require that
a transaction already exists before being called. The doTheUpdate()
method can really have it either way, as long as there is an active
transaction associated to the thread calling it.
regards,
Patrik Andersson
-Ursprungligt meddelande-
Från: Joni Suominen [mailto:[EMAIL PROTECTED]]
Skickat: den 6 juli 2001 08:32
Till: Orion-Interest
Ämne: LDAP and EJBs
Hi everyone!
I have a use case where I need to interface an LDAP directory and a
relational database within a same transaction. That is, to create a
new
row to RDBMS from session bean using entity bean and to create an LDAP
entry from the same session bean using Sun's LDAP provider through
JNDI.
1. Is this possible?
2. Can I use container managed transactions?
Thanks a lot!
--
Joni
[EMAIL PROTECTED]
Startup script for HP-UX
Hi! Does any of you have a proper Orion startup script for HP-UX? I found a one from Orion support which works fine in Linux but not in HP-UX. I also have problems to shutdown Orion in HP-UX. Nothing happens when I issue: /opt/java1.3/bin/java -jar admin.jar ormi://localhost admin passwd -shutdown force -- Joni [EMAIL PROTECTED]
LDAP and EJBs
Hi everyone! I have a use case where I need to interface an LDAP directory and a relational database within a same transaction. That is, to create a new row to RDBMS from session bean using entity bean and to create an LDAP entry from the same session bean using Sun's LDAP provider through JNDI. 1. Is this possible? 2. Can I use container managed transactions? Thanks a lot! -- Joni [EMAIL PROTECTED]
Re: Debugging and orion
Hi, You can use -out and -err commandline swiches to specify where System.outs and System.errs are written. For example: $JAVA_HOME/bin/java -server -jar orion.jar \ -config $ORION_HOME/config/server.xml \ -out $ORION_HOME/log/server-out.log \ -err $ORION_HOME/log/server-err.log -- Joni [EMAIL PROTECTED] Nusairat, Joseph F. wrote: OK its an older thread ... but i dont have time to read them all I was just wondering ... if i run Orion as a service ... how do i see the System.out.printlns? Does it write them to any log file?? OR is there anyway i can pick them up -Original Message- From: Justen Stepka [mailto:[EMAIL PROTECTED]] Sent: Wednesday, February 21, 2001 11:08 AM To: Orion-Interest Subject: Re: Debugging and orion No matter what anyone tells you, System.out.println (ObjectName:MethodName ():line number); will work wonders! Justen Stepka On Wed, 21 Feb 2001, Julian Richardson wrote: Hi, What tools / IDEs are people out there using to develop EJBs for use with Orion? At the moment we (as a project team, rather than company) are trying to standardise on an app server and IDE to use, requirements being the usual: cost speed flexibility resource usage debugging support from the IDE - I saw that someone mentioned using Kawa with Orion on the orionsupport website; has this combination been used by many people? (I'm new to the app server world, although I've been doing server-side Java work for several years now - I'm not used to environments where it's even possible to use an integrated debugger, but others on the team say it's necessary :-) Comments about the speed / flexibility / resource usage of Orion are welcome too, from the website it looks pretty slick and the documentation looks good, but time will tell... also if there's any major snags with orion in terms of things that the current version doesn't do correctly or doesn't implement... cheers Jules --
Different JNDI context factories
Hi! I made an experiment with a setup where ejb components are deployed in another server than web components. Orion provides three types of JNDI context factories: com.evermind.server.rmi/RMIInitialContextFactory com.evermind.server.ApplicationClientInitialContextFactory com.evermind.server.ApplicationInitialContextFactory ApplicationClientInitialContextFactory is meant for Java application clients but what is the difference between the other two, RMIInitialContextFactory and ApplicationInitialContextFactory? My experiment worked just fine no matter which one I used when separating web tier from ejb tier. -- Joni [EMAIL PROTECTED]
Re: Using JAAS for authentication (Was: bugs in @page extends=...)
this helps! Anyway, since you already have your own implementation
which works, you don't have any hurry to start using JAAS. If I were you
I would swich to JAAS at the same time when you are considering to swich
to JDK 1.4.
--
Joni
[EMAIL PROTECTED]
Just curious to see how I could use this standard API without losing all
the flexibility and dynamicity that I've already accomplished with my
own implementation.
Regards and thanks for the info,
D.
Joni Suominen wrote:
Hi Daniel,
JAAS is not necessarily tied to the OS user. Actually you can tie it to
the OS user by using proper login modules which can authenticate if a
user is already logged into an OS. However, in a true Java spirit, JAAS
is much more generic. In fact it is just a framework to implement
versatile authentication and authorization schemes. For instance, I
authenticate user's against a relational database (some might use LDAP).
The JAAS provides abstractions to represent user's identity and
permissions. It also provides algortihms to check if configured security
policy implies certain permission.
Some sample code:
PagePermission pagePermission = new
PagePermission(admin);
PermissionCollection pc =
Policy.getPolicy().getPermissions(user.getSubject(), null);
if (pc.implies(pagePermission)) {
// authorization succeeded...
}
else {
// authorization failed, the current user don't have a permission to
view a page on this domain.
}
Then on policy file I might have:
grant Principal org.shiftctrl.framework.security.SCGroupPrincipal
admin {
permission
org.shiftctrl.framework.security.permissions.PagePermission admin;
};
This way it is easy to implement multiuser Java applications where
accesses are controlled on per-user or per-group level. JAAS is also
integrated to the JDK 1.4, like the new logging API you mentioned.
--
Joni
[EMAIL PROTECTED]
snipped...
Re: bugs in @page extends=.... in Orion implementation?
Hi Daniel,
JAAS is not necessarily tied to the OS user. Actually you can tie it to
the OS user by using proper login modules which can authenticate if a
user is already logged into an OS. However, in a true Java spirit, JAAS
is much more generic. In fact it is just a framework to implement
versatile authentication and authorization schemes. For instance, I
authenticate user's against a relational database (some might use LDAP).
The JAAS provides abstractions to represent user's identity and
permissions. It also provides algortihms to check if configured security
policy implies certain permission.
Some sample code:
PagePermission pagePermission = new
PagePermission(admin);
PermissionCollection pc =
Policy.getPolicy().getPermissions(user.getSubject(), null);
if (pc.implies(pagePermission)) {
// authorization succeeded...
}
else {
// authorization failed, the current user don't have a permission to
view a page on this domain.
}
Then on policy file I might have:
grant Principal org.shiftctrl.framework.security.SCGroupPrincipal
admin {
permission
org.shiftctrl.framework.security.permissions.PagePermission admin;
};
This way it is easy to implement multiuser Java applications where
accesses are controlled on per-user or per-group level. JAAS is also
integrated to the JDK 1.4, like the new logging API you mentioned.
--
Joni
[EMAIL PROTECTED]
Daniel López wrote:
Hi Joni,
I agree it can be done. My situation is that I just don't log requests
but I can log everything that happens in the environment, security
settings, internationalization issues, XML/XSLT performance... so my
logging code is tightly integrated with the framework and I can control
it from a single point. I could use a filter to check things before the
request is served, but as the rest of the parts are going to use it,
what do I win by externalising it. Anyway, I hope to be able to discard
this part of the environment as soon as JDK1.4 logging features become a
reality. With security is more or less the same, we implemented our own
security library because we are not satisfied with the standard and we
don't want to get container specific. I agree I could externalise this
as a filter but I wouldn't really win much, as it is also quite tightly
integrated with the rest of the things. IMO, things that are standard
and can be switched on/off and replaced with another implementation
without a hitch are candidates to be externalised, otherwise, it's not
that it cannot be done with filters, it's that I don't see a big win in
doing so. Of course, that's my situation because I can control the MVC 2
implementation and I can tightly integrate these parts.
Just a question. I thought JAAS was used to control if the OS user that
was executing the Java code was/was not allowed to perform such
execution. If the OS user that executes the servlet code is supposedly
always the same, how is this thing related to the application user that
performs the request? I didn't think this standard could be used for
that, if it is, I might go for it ;).
regards,
Dan
Joni Suominen wrote:
Hi!
I have implemented these features as externalized filters. The reason
for this was that I use a MVC 2 product which is developed independently
(Jakarta Struts http://jakarta.apache.org/struts/). Like Marcel said the
beauty of filters is that they are independent from the rest of the
application. The filters allow me to plug in these services to the
application without any integration code. So, I have a LogFilter which
logs the requests using log4j (http://jakarta.apache.org/log4j/). After
that there's an AuthorizationFilter which checks if the current user has
priviledges to execute the request. AuthorizationFilter utilizes the
forthcoming Java Authentication and Authorization service
(http://java.sun.com/products/jaas/). All in all, I have a pretty
flexible system with a very little own code.
Joni
[EMAIL PROTECTED]
Daniel López wrote:
Hi,
I've also used filters, but in this case to filter requests by IP, as I
don't want certain IP's to be able to access certain URLs. On the other
hand, I implemented security, logging and other features inside my own
servlet Controller, following a Model 2 approach. I thought about
externalizing this services as filters, but that would mean that the
web.xml file will grow too much and that I would have to put too much
things in the request object each time. It's not that I've discarded the
idea, but I don't see any compelling reason to implement everything as
filters. We don't use servlets, but ONE controller servlet and action
classes and we use JSP as view-only, creatures (actually, we seldom use
them as we use XML/XSLT), so as anyway all the requests go through my
controller servlet... We implemented IP filtering as a filter because we
also wanted to use it in other
Re: Writing an external client
Hi Kevin! Please check that applications\CricetApp is in your CLASSPATH. -- Joni [EMAIL PROTECTED] Kevin Jones wrote: I want to access a javax.sql.DataSource from an external client, or, failing that an EKB Session bean. Looking at the (scant) Orion docs I have to do something like this Hashtable env = new java.util.Hashtable(); env.put(Context.INITIAL_CONTEXT_FACTORY,com.evermind.server.ApplicationClie ntInitialContextFactory); env.put(Context.PROVIDER_URL,ormi://localhost/CricketApp); Context initial = new InitialContext(env); However I get an exception at the call to new InitialContext(env) javax.naming.NamingException: META-INF/application-client.xml resource not found at com.evermind.server.ApplicationClientInitialContextFactory.getInitialContext (JAX) at javax.naming.spi.NamingManager.getInitialContext(NamingManager.java:665) at javax.naming.InitialContext.getDefaultInitCtx(InitialContext.java:246) at javax.naming.InitialContext.init(InitialContext.java:222) at javax.naming.InitialContext.init(InitialContext.java:198) at SimpleClient.main(SimpleClient.java:18) I've put an application-client.xml in applications\CricketApp\META-INF, but, obviously it's not being found. What am I missing/doing wrong? Kevin Jones DevelopMentor www.develop.com
Re: bugs in @page extends=.... in Orion implementation?
Hi! I have implemented these features as externalized filters. The reason for this was that I use a MVC 2 product which is developed independently (Jakarta Struts http://jakarta.apache.org/struts/). Like Marcel said the beauty of filters is that they are independent from the rest of the application. The filters allow me to plug in these services to the application without any integration code. So, I have a LogFilter which logs the requests using log4j (http://jakarta.apache.org/log4j/). After that there's an AuthorizationFilter which checks if the current user has priviledges to execute the request. AuthorizationFilter utilizes the forthcoming Java Authentication and Authorization service (http://java.sun.com/products/jaas/). All in all, I have a pretty flexible system with a very little own code. Joni [EMAIL PROTECTED] Daniel López wrote: Hi, I've also used filters, but in this case to filter requests by IP, as I don't want certain IP's to be able to access certain URLs. On the other hand, I implemented security, logging and other features inside my own servlet Controller, following a Model 2 approach. I thought about externalizing this services as filters, but that would mean that the web.xml file will grow too much and that I would have to put too much things in the request object each time. It's not that I've discarded the idea, but I don't see any compelling reason to implement everything as filters. We don't use servlets, but ONE controller servlet and action classes and we use JSP as view-only, creatures (actually, we seldom use them as we use XML/XSLT), so as anyway all the requests go through my controller servlet... We implemented IP filtering as a filter because we also wanted to use it in other applications, as well as with static content. Just my 2c, D. --- Daniel Lopez Janariz ([EMAIL PROTECTED]) Web Services Computer Center Balearic Islands University --- Marcel Schutte wrote: Hi Randahl, I've used a filter to do gzip encoding for specific file types. Starting with the tutorial at http://www.orionserver.com/tutorials/filters/lesson5/, I've changed some things to get the hang of it and also to reduce the amount of buffering used. It works fine, haven't had a problem with this (intranet)site. The beauty of filters is that they are completely independent from your application, just edit your web.xml and you are set. Let me know if you wnat some more info. Marcel -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of Randahl Fink Isaksen Sent: Wednesday, May 30, 2001 9:39 AM To: Orion-Interest Subject: RE: bugs in @page extends= in Orion implementation? Hi Marcel I looked it up, and I definately see your point about the filter mechanism. Have you any experience with this on Orion? Does it work well? And what have you used it for? Of course if anyone else have tried it out, I would very much like to hear from you too. Yours Randahl -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of Marcel Schutte Sent: 30. maj 2001 00:22 To: Orion-Interest Subject: Re: bugs in @page extends= in Orion implementation? That's right, the two examples Randahl gives can be easily solved using the servlet2.3 filter mechanism, giving a very loose coupling of these features with your actual application. Marcel - Original Message - From: Juan Lorandi (Chile) [EMAIL PROTECTED] To: Orion-Interest [EMAIL PROTECTED] Sent: Tuesday, May 29, 2001 10:45 PM Subject: RE: bugs in @page extends= in Orion implementation? I believe this could be better modeled differently, because it leaves no room for the web server to implement its callbacks and java doesn't support multiple inheritance. -Original Message- From: Randahl Fink Isaksen [mailto:[EMAIL PROTECTED]] Sent: Martes, 29 de Mayo de 2001 3:01 To: Orion-Interest Subject: RE: bugs in @page extends= in Orion implementation? Well, can anyone confirm that JSP pages can extend your own class when using Orion??? Or has noone been able to use @page extends? I see many good reasons for implementing your own super class for your JSP pages. For instance, such a superclass could do logging of all client activities or it could do security checking. Randahl -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of Robert Nicholson Sent: 25. maj 2001 21:15 To: Orion-Interest Subject: RE: bugs in @page extends= in Orion implementation? It's doing you a favour.Please elaborate on why you want
Re: finderException in home interface of CMP ??
In theory a multi-object finder can throw an exception, at least when using BMP. This is possible for instance when a multi-object finder calls a single-object finder. I am not sure if it would be a good programming practise to let this FinderException to go unchecked to a caller though. -- Joni [EMAIL PROTECTED] Eddie wrote: I understand your point and I completely agree that it's illogical. That's why I started testing around and found out that it's not necessary to throw a finderexception in case of a return collection. Like I see it now: I finderException is never thrown in case of a collection as it always returns an empty collection when it can't find anything ? Or am I missing something here. Neither does Orion complain when you leave out the FinderException! Or are there other situations when a FinderExceptions can occur in case of a Collection ?? Eddie - Original Message - From: Joni Suominen [EMAIL PROTECTED] To: Orion-Interest [EMAIL PROTECTED] Sent: Monday, May 21, 2001 6:41 PM Subject: Re: finderException in home interface of CMP ?? Joni Suominen wrote: Eddie wrote: To me it's a bit unclear when I need to throw a FinderException in the finder methods in the Home interface with CMP. ? Now I only throw a finderException in case of - the findByPrimareyKey (otherwise orion complaints), - and in case of a single return object (orion doesn't complain, but gives strange runtime errors). So when a finder returns a collection I don't throw a finderException!! Is this correct ? No. Every finder method should declare that it throws FinderException. Joni [EMAIL PROTECTED] To be more specific: In the home interface you should declare all the fonder methods to throw FinderException. But when a method which returns a non-collection raises an exception it will be ObjectNotFoundException (which is a subclass of FinderException). However, if a finder method which returns a collection don't find any matching objects an empty collection is returned. The method signature must still declare the FinderException in home interface which seems to be a bit illogical? Joni [EMAIL PROTECTED]
Re: finderException in home interface of CMP ??
Joni Suominen wrote: Eddie wrote: To me it's a bit unclear when I need to throw a FinderException in the finder methods in the Home interface with CMP. ? Now I only throw a finderException in case of - the findByPrimareyKey (otherwise orion complaints), - and in case of a single return object (orion doesn't complain, but gives strange runtime errors). So when a finder returns a collection I don't throw a finderException!! Is this correct ? No. Every finder method should declare that it throws FinderException. Joni [EMAIL PROTECTED] To be more specific: In the home interface you should declare all the fonder methods to throw FinderException. But when a method which returns a non-collection raises an exception it will be ObjectNotFoundException (which is a subclass of FinderException). However, if a finder method which returns a collection don't find any matching objects an empty collection is returned. The method signature must still declare the FinderException in home interface which seems to be a bit illogical? Joni [EMAIL PROTECTED]
Load balancing advice needed
Hi! I am currently planning the new production environment for our service. The idea is to have following harware setup: 1 firewall (which used for loadbalancing too) 2 Linux/Intel boxes as frontend servers (both running HTTP accelerators and web containers) 1 UNIX server as a backend server (running an EJB container and databases) I went through load balancing documentation which I found from orionserver.com but I didn't find any way to separate Orion's web container and EJB container to achieve the above setup. Is it somehow possible to do that kind of separation? How would you distribute the required processes on those three servers? -- Joni [EMAIL PROTECTED]
Re: Generating primary keys
Hi!
I use a bit similar approach that in the Brett's article. The difference
is that instead of hitting database every time a small cache is kept for
each sequence. The algorithm is simple. Each unique id consists of two
values: high and low (uid = high + low). The UidGenerator is initialized
from a database. It exclusively reads a high value from db and updates
high value in database by one. This ensures that each possible generator
gets a unique high value. Then on each getNextUid() call a value high +
low is returned and low is increased by one. If low value hits its
maximum it is set back to 0 and a new high value is read from the
database.
Consequences:
+ better scalability since most of the uids are generated in memory
without database access
+ works also in a distributed environment since each UidGenerator is
initialized with a unique high value
+ should work on any J2EE app server
- leaves unused uids every time the system is booted
This HIGH-LOW algorithm is explained in:
http://www.sdmagazine.com/articles/1999/9912/9912p/9912p.htm
Here's some sample code. My version is implemented as a BMP entity bean
and uids are Java long variables. LOW cache is set to be 16 numbers but
this is easy to change just by tweaking the constants: LOW_LENGTH and
LOW_MAX.
UidGenerator.java:
##
package org.shiftctrl.uid.persistence;
import java.rmi.RemoteException;
import javax.ejb.EJBObject;
/*
* @author Joni Suominen
* @version
*/
public interface UidGenerator extends EJBObject {
public long getUid() throws RemoteException;
}
UidGeneratorHome.java
#
package org.shiftctrl.uid.persistence;
import java.rmi.RemoteException;
import javax.ejb.CreateException;
import javax.ejb.FinderException;
import javax.ejb.EJBHome;
import javax.naming.NamingException;
/**
* The home interface of the UidGenerator EJB.
*/
public interface UidGeneratorHome extends EJBHome {
public UidGenerator create(String id)
throws RemoteException, NamingException, CreateException;
public UidGenerator findByPrimaryKey(String id)
throws RemoteException, FinderException;
}
UidGeneratorEJB.java
package org.shiftctrl.uid.persistence;
import java.sql.Connection;
import java.sql.PreparedStatement;
import java.sql.ResultSet;
import java.sql.SQLException;
import javax.ejb.EntityBean;
import javax.ejb.EntityContext;
import javax.ejb.EJBException;
import javax.ejb.CreateException;
import javax.ejb.FinderException;
import javax.ejb.ObjectNotFoundException;
import javax.naming.InitialContext;
import javax.naming.NamingException;
import javax.sql.DataSource;
import org.shiftctrl.framework.util.JndiNames;
/**
* A unique id generator implementing
* a
href=http://www.sdmagazine.com/articles/1999/9912/9912p/9912p.htm;
* high-low algorithm/a using BMP on relational database.
* p
* The required database structure:br
* CREATE TABLE uids (id VARCHAR PRIMARY KEY, high LONGINT);
*
* p
* The usage:
* pre
*
* UidGenerator uidGenerator = null;
* try {
* uidGenerator = uidGeneratorHome.findByPrimaryKey(Generator
name);
* } catch (FinderException fe) {
* try {
* System.out.println(No existing UidGenerator \Generator
name\, let's create a one);
* uidGeneratorHome.create(Generator name);
* uidGenerator = uidGeneratorHome.findByPrimaryKey(Generator
name);
* } catch (Exception e) {
* throw new EJBException(e);
* }
* }
*
* uidGenerator.getId();
*
* /pre
*
* @author Joni Suominen
* @version $Revision: 1.2 $
*/
public class UidGeneratorEJB implements EntityBean {
/**
* An id for a generator.
*/
private String id;
/**
* The high value which forms the most significant bits of id.
*/
private long high;
private static final int HIGH_LENGTH = 60; // 60 bits
/**
* The low value which forms the least significant bits of id.
*/
private int low;
private static final int LOW_LENGTH = 4; // 4 bits
private static final int LOW_MAX = 15;// 2^4 - 1
private EntityContext context;
/**
* Creates an UidGenerator.
*
* @param id an id for UidGenerator
*/
public String ejbCreate(String id)
throws NamingException, CreateException {
this.id = id;
this.high = 1;
this.low = 0;
Connection con = null;
PreparedStatement pStmt = null;
try {
con = getConnection();
pStmt = con.prepareStatement(insert into uids (id, high)
values (?, ?));
pStmt.setString(1, this.id);
pStmt.setLong(2, this.high + 1);
pStmt.executeUpdate();
pStmt.close();
} catch (SQLException se) {
se.printStackTrace();
throw new EJBException(se);
} finally {
if (pStmt != null) {
try {
pStmt.close
Problems with SAPDB
Hi! I am in a middle of changing our database from HypersonicSQL to SAPDB. In theory it should be easy ;) This is how I have proceeded: 1. I created a database-schema for SAPDB (actually I found a one from this mailing list and copied it). I put that file to $ORION_DIR/config/database-schemas/. The schema file being used: ?xml version=1.0? !DOCTYPE database-schema PUBLIC -//Evermind//- Database schema http://www.orionserver.com/dtds/database-schemas.dtd; database-schema name=SapDB not-null=not null null= primary-key=primary key max-table-name-length=32 type-mapping type=boolean name=boolean / type-mapping type=java.lang.Booleanname=boolean / type-mapping type=byte name=byte / type-mapping type=java.lang.Byte name=byte / type-mapping type=char name=char / type-mapping type=java.lang.Character name=char / type-mapping type=double name=double precision / type-mapping type=java.lang.Double name=double precision / type-mapping type=floatname=float / type-mapping type=java.lang.Float name=float / type-mapping type=int name=integer / type-mapping type=java.lang.Integername=integer / type-mapping type=long name=integer / type-mapping type=java.lang.Long name=integer / type-mapping type=shortname=smallint / type-mapping type=java.lang.Short name=smallint / type-mapping type=java.lang.String name=varchar(255) / type-mapping type=java.math.BigDecimal name=fixed(15,15) / type-mapping type=byte[] name=long byte / type-mapping type=java.sql.Datename=date / type-mapping type=java.sql.Timename=time / type-mapping type=java.sql.Timestamp name=timestamp / type-mapping type=java.util.Date name=timestamp / type-mapping type=java.io.Serializable name=long byte / disallowed-field name=key / disallowed-field name=date / disallowed-field name=timestamp / disallowed-field name=time / disallowed-field name=username / disallowed-field name=password / disallowed-field name=order / disallowed-field name=table / disallowed-field name=value / disallowed-field name=name / disallowed-field name=count / /database-schema 2. I created a data-source description for SAPDB and pointed my application to use it. The data-source description being used: data-sources data-source class=com.evermind.sql.ConnectionDataSource inactivity-timeout=120 location=jdbc/DefaultCoreDS pooled-location=jdbc/DefaultPooledDS name=SAPTest data-source schema=./database-schemas/sapdb.xml xa-location=jdbc/xa/DefaultXADS ejb-location=jdbc/DefaultDS url=jdbc:sapdb://localhost/testdb connection-driver=com.sap.dbtech.jdbc.DriverSapDB username=TEST password=TEST / /data-sources 3. I added the SAPDB JDBC driver to $ORION_DIR/lib 4. I restarted Orion I believe that in an ideal case these would be the required steps when database is changed. However, I encountered two problems: 1. Table auto-creation was not fully successful. All the other tables were correctly created but one. SAPDB has a limit for maximum lenght for table row names but I didn't find any way to specify this in schema definition. Maximum legth for table names are supported by the attribute max-table-name-length=32 and it works fine. Do you know if there's a similar way to restrict the length of a row? (I tried max-row-name-length but wasn't lucky :) 2. The following exception is thrown: com.evermind.server.rmi.OrionRemoteException: javax.ejb.EJBException: Error saving state: Connection was closed at com.evermind.server.ejb.EJBUtils.getUserException(JAX) at UserManager_StatelessSessionBeanWrapper14.getGroupNamesForUser(UserManager_StatelessSessionBeanWrapper14.java:1348) at org.shiftctrl.framework.security.authentication.DefaultLoginModule.commit(DefaultLoginModule.java:243) ... Nested exception is: javax.ejb.EJBException: Error saving state: Connection was closed at SCUser_EntityBeanWrapper5.saveState(SCUser_EntityBeanWrapper5.java:1057) at SCUser_EntityBeanWrapper5.getGroupNames(SCUser_EntityBeanWrapper5.java:712) at org.shiftctrl.user_management.service.UserManagerEJB.getGroupNamesForUser(UserManagerEJB.java:328) ... The control goes so that the user is first authenticated using a session bean + CMP entity bean. Then the user's login metadata (last login, num of logins et.) is correctly updated using a session bean + CMP entity bean. After that the intention is to initialize user's session by calling getGroupNamesForUser() on a session
Re: Problems with SAPDB (correction)
I meant column names when I wrote row names. As far as I know the row's usually don't have any names ;) Stupid me... Anyway, the problems remain... Joni [EMAIL PROTECTED]
Stateless bean and remove
Hi! This question should propably be asked in EJB-INTEREST mailing-list but I raise it here since I am not currently following it. So apologize me if you feel that this is a bit offtopic. The question is simple: Is it required to call the remove() method on stateless session bean after finishing using it or can the container automatically restore it to a pool (using Java garbage collection or DGC perhaps?)? For instance, the ATM example which ships with Orion don't use remove() after finishing the use of a stateless session bean instance. The same goes with Sun's Java Pet Store demo. On the other hand in Wrox Press's book Professional Java Server Programming, J2EE Edition they call remove() on stateless session beans. Currently I don't use remove() with stateless session beans. Should I? -- Joni [EMAIL PROTECTED]
Re: Standar Template
Using a filter sounds like a good idea. It is a natural way to implement Decorator design pattern. -- Joni [EMAIL PROTECTED] Smith Jason wrote: Maybe you could use a filter? Check out the filter tutorial at http://www.orionserver.com/ /Jason -Original Message- From: Dave Ford [mailto:[EMAIL PROTECTED]] Sent: Thursday, May 10, 2001 9:17 PM To: Orion-Interest Cc: Dan Tharp Subject: Standar Template I want to create a web app in which every page on the site has a standard header along the top and a standard menu along the left edge (a pretty standard thing). I came up with 2 ways of doing this: 1. Use a table tag and jsp:include tags on EVERY page: table tr tdjsp:include page=standardHeader.jsp//td /tr tr td colspan=2 table tr td valign=topjsp:include page=/menu.jsp //td td valign=top THIS IS WHERE THE PAGE-SPECIFIC CONTENT (i.e. the body)* /td /tr /table /td /tr /table 2. Invert the above solution to create one master template (or controller) and have the content page name passed in as a parameter. Here would be the master template-controller page: table tr tdjsp:include page=standardHeader.jsp//td /tr tr td colspan=2 table tr td valign=topjsp:include page=/menu.jsp //td td valign=top jsp:include page=%=request.getParameter(contentPage)% /* /td /tr /table /td /tr /table The key difference between these two architectures are best understood by looking at the 2 lines with the * at the end. Also, in option 2, there is only one copy of the above code. In option 1, there is one copy per content page Q1: Does anyone have any preference between options 1 and 2? Q2: Is there a better way of achieving this result? Q3: Do either have any negetive drawback I need to consider? (I will be converting an entire site) By the way, I'm currently achieving this effect VERY easily using good old client-side html frames. But due to popular demand, framse must go. Dave Ford Smart Soft - The Java Training Company http://www.smart-soft.com
Re: Servlet Mapping Recursion Problem
Hi Bill! I am using a front servlet pattern with Orion. I use Struts but I suppose this approach should work in any framework. 1. The front servlet mapping in web.xml: servlet-mapping servlet-nameaction/servlet-name url-pattern*.html/url-pattern /servlet-mapping So all the requests ending to .html are processed by the front-end servlet. 2. All the valid URL's are mapped in an xml config file. The front servlet initializes its state by reading this file in a startup. An entry in this file looks like: actionpath=/faq type=org.shiftctrl.general.interaction.EmptyAction forward name=success path=/faq.jsp/ /action This means that the request http://myserver.com/faq.html is first processed by front servlet. The servlet searches its configured mapping and forwards the request. In this case it would forward the request to faq.jsp. I haven't had problems with this setup and I really like it for following reasons: 1. It is easy pipeline actions. Just configure the mapping so that its forward part points to another mapping. 2. You can hide the technology being used. Clients only see URLs which are always ending .html (so .jsps are not exposed outside, their function is only to provide the view part in MVC pattern) 3. It is possible to dynamically change the URL mapping by calling appropriate methods in front servlet. -- Joni [EMAIL PROTECTED] Bill Winspur wrote: In my web.xml I set up the servlet mapping, for a (front) servlet to handle all requests for the web app, to '/'. This works to the extent that all requests are passed to the servlet. BUT... if I do a forward to another resource from the servlet, the servlet's doGet() method is invoked recursively with no response generated until the jvm runs out of memory. The same thing happens with a url-pattern of '/*'. If the servlet itself generates html output, that works. If I map my servlet to a specific resource, like /index.html, forwarding works fine, but the servlet cant handle all requests and 'http://myhost/myapp' returns 'no directory browsing'. The url-pattern '/' seems like a valid directory specification (as per the servlet 2.2 spec, 'Servlet Mapping Techniques') and a front component, handling all requests is fundamental in the j2EE blueprints (http://java.sun.com/j2ee/blueprints/index.html, section 10.6), so my approach seems architecturally reasonable. I've looked thru the archives and found two reports of the same problem. A proposed workaround from Magnus seems to recommend explicitly mapping every request your app generates (which would make the web.xml coding difficult to maintain in a large app) pendng the delivery of 'filters', perhaps in Servlet 2.3. I dont really understand what Magnus is suggesting: PS: If you're bound to 2.2 or lower then you will not be able to use Filters, in that case you'd have to apply a more complex solution with duplicate resource roots and so on. It looks like 'no front servlets on Orion' to me, but I hope I'm wrong. I was not able to find any references to the problem on bugzilla, but I'm not very adept with its query interface. My hunch is that Orion's request dispatcher is invoked in two contexts (from a client, and from a forward) and does not distinguish between them when it comes to selecting a resource to respond to the client. If the dispatcher ignored mapping in the forward context and just accepted the specified resource, the recursion problem would vanish and the front servlet would trap ALL client requests as per the blueprint architecture. Just a hunch, things are probably not that simple. Does anyone know if this is a problem with other webservers ? My experience with Apache was all static pages. My environment is NT4(sp6), Orion 1.4.8, IE 5.00, JDK 1.2.2. Also, if somebody is actually using the front servlet pattern and univeral mapping on Orion, and could let me know how to set it up and/or get around the recursion I'd be a happy camper. TIA, Bill.
Re: FW: custom finder in CMPs (SLSB facade)
Armin Michel wrote:
IMHO, facades are only useful when they do some extra-work, e.g.
consolidating the work with many other EJBs.
When you just use a facade to plainly forward any request to exactly one
EntityBean (1:1 relationship between facades and EntityBeans) than it's not
worth the trouble.
Except if there's a need to keep the design uniform. I am always using a
following layered architecture in my J2EE applications:
+---+---+ +---+---+---+
| P | I | | S | D | P |
| r | n | | e | o | e |
| e | t | | r | m | r |
| s | e | | v | a | s |
| e | r | | i | i | i |
| n | a | | c | n | s |
| t | c | | e | | t |
| a | t | | | | e |
| t | i | | | | |
| i | o | | | | |
| o | n | | | | |
| n | | | | | |
+---+---+--+---+---+---+
|Framework |
+--+
Presentation and interaction layers are deployed on web container.
Presentation layer contains HTML, WML or XML specific code (e.g. JSP tag
extensions). Interaction layer contains MVC command classes and
presentation related model components. The rest of the tiers, service,
domain and persistence are deployed on EJB container. Service tier,
implemented as session beans, provide the use case implementations and
serves them as a facade to the interaction tier. Domain layer contains
CMP entity beans and dependent objects. Persistency tier contains
possible data source specific code. For instance BMP entity beans. The
framework package contains some common classes used everywhere. I
usually deploy the framework jar file to ${JAVA_HOME}/jre/lib/ext.
The good thing about this architecture is that it is easy understand
even complex use case realizations since every use case follow exactly
the same design. The bad thing is that it requires some extra code (and
an extra EJB call) when doing simple things (like mentioned in previous
mails). But I later hope to resolve this extra programming overhead by
using code generators.
--
Joni
[EMAIL PROTECTED]
Re: remote shutdown of orion 1.4.8
[EMAIL PROTECTED] wrote: Folks, The following remote shutdown command does not work anymore in 1.4.8 (it works fine in 1.4.7), any ideas? bug? java -jar admin.jar ormi://localhost admin password -shutdown Hi! It is a bit unclear for me how the shutdown process should work. When I initiate java -jar admin.jar ormi://localhost admin password -shutdown the message Shutting down... is printed on the log. After that Orion no longer responds to any requests but all the Java processes are still kicking alive: $ ps ax 21242 pts/2S 0:00 /usr/java/jdk1.3/bin/i386/native_threads/java -server -Xms8m -Xmx192m -Djava... 21266 pts/2S 0:00 /usr/java/jdk1.3/bin/i386/native_threads/java -server -Xms8m -Xmx192m -Djava... ... Platform: RedHat Linux 7.0, HotSpot Virtual Machine, Orion 1.4.7 Is there a way to shutdown Orion so that it stops all Java procesesses too? -- Joni [EMAIL PROTECTED]
Re: JavaMail
Nijhawan, Sumit wrote: Hi, Does anyone know of any products (other than Orion) using and being distributed with Sun's reference implementation of the JavaMail API (mail.jar and activation.jar)? Is anyone aware of any licensing issues for distributing these jars freely? Thanks. From JavaMail FAQ (http://java.sun.com/products/javamail/FAQ.html): Q: Is the JavaMail API implementation completely free? Can I ship it along with my product? A: Yes. The current release of the JavaMail API implementation, is completely free and you can include it in your product. This release includes IMAP, POP3, and SMTP providers as well. Please do read the LICENSE and ensure that you understand it. The JavaBeans Activation Framework is also free for use under a similar license. -- Joni [EMAIL PROTECTED]
Re: Is it possible to have Apache as Reverse Proxy (front-end) and LoadBalancer ?
Take a look at Squid (http://www.squid-cache.org). I have successfully used it as a front-end accelerator with Orion. With simple redirect rules I guess you can use it as a load balancer too. It is open source, very stable and available for most platforms. Although I found a possible bug in Orion relating to frontend tag. See: http://www.mail-archive.com/orion-interest%40orionserver.com/msg12780.html There's also Linux Virtual Server project which is aimed for load balancing: http://www.linuxvirtualserver.org/ Joni [EMAIL PROTECTED] Joseph B. Ottinger wrote: Sure, it's possible. Dumb, but possible. See www.orionsupport.com for more on it. Note that we saw degradation on a massive scale - something that took 12 seconds without apache proxying took 1 minute 45 seconds with apache, in the worst case. :) On Mon, Apr 30, 2001 at 08:04:17PM +0200, Ismael wrote: Is it possible to have Apache as Reverse Proxy (front-end) and LoadBalancer at the same time.? I think that on the servers you define the front-end and they will collide. Anyone knows how to do it? -- --- Joseph B. Ottinger [EMAIL PROTECTED] http://epesh.com/ IT Consultant
Re: MVC/XML Framework Comments please
Cory Updyke wrote: My issue with XML/XSL transformations is this: My theory for the presentation tier is to keep it as simple as possible. Occasionally, you find a designer who has every right to be a engineer, but designs because they enjoy it. This person will scoff(sp?) at the simplicity of XSL. Most of the time you will find designers who know HTML (PDF, etc.). Which concept is closer to HTML, XML/XSL or JSP? I think JSP. I also have found it easier to say Do you see the green (or whatever color depending upon your IDE) colored text? Please don't delete that than it is to teach someone the concept behind XSL. You can have a good hour conversation, hand them the JSP Syntax Reference document and say DESIGN good man. I totally agree! Some years ago (before JSP) we implemented an XML based publishing system. We used our own template language which was a simplified version of XSL. The platform was techinically excellent and contained more than 150.000 lines of Java code. We had a very complex layered caching mechanism: cache for data objects, cache for XML data sources, cache for precompiled templates, cache for generated HTML snippets etc. The bad news for us came later. The system was a way too complex for our designers and HTML coders! It took relatively long time to do simple things. In other words, we went too far with the abstraction. Anyway... enough talk... we use Struts and our own Model-2-Brew. That's what we do also nowdays. We use Struts with some own add ons which allow us to configure the caching policies and access control lists for each action in struts-config.xml. Business objects and processes are without an exception implemented as EJBs. Joni [EMAIL PROTECTED]
sendRedirect and frontend host=.. port=.. / bug
Hi! There seems to be a bug in Orion (1.4.7) with sendRedirect and frontend host=.. port=.. / tag (from web-site.xml). I have understood the meaning of frontend tag so that it is used in HTTP REDIRECTS if relative URL is used in a method HttpServletResponse.sendRedirect. E.g. response.sendRedirect(/foo.jsp) should send a redirect pointing to http://my.server:port/foo.jsp if web-site is configured with frontend host=my.server port=port / tag. This doesn't work in my environment where I use Squid cache server (http://www.squid-cache.org) as a frontend server listening port 80 and Orion is running on a same machine on port 8080. So, my web-site configuration is: frontend host=www.myserver.com port=80 / The port attribute in frontend tag is working well but Orion always incorrectly sends the redirect pointing to localhost, thus neglecting the host attribute. Notice that the actual request comes from localhost since Squid is listening port 80 and then issueing a new HTTP request to port 80. Here's a beautiful picture depicting this architecture :) +-+ | Firewall | +-+ | | http://www.myserver.com:80 V +---+ | ++ | | | Squid| | | ++ | | || | | http://localhost:8080 | | V| | ++ | | | Orion| | | ++ | | | +---+ The obvious (but ugly) fix for this is to always use absolute URL's when using response.sendRedirect(); This bug propably have implications on a various load-balancing solutions too? Has any of you seen this? I am considering to put this to Orion bugzilla but I am still hesitating since I am not 100% sure if I have understood the problem correctly. BTW. This Squid set up as a front-end cache server is an excellent way to speed up Orion. It is very fast with static files (gifs, jpegs etc.) and with a simple rules + HTTP headers it is possible to cache dynamic pages too. You can contact me in case you want more information. -- Joni [EMAIL PROTECTED]
