RE: [OT] Clustering
I assume you ask about an Oracle RAC database. In that case, the mode is "kind-of" ACTIVE/ACTIVE: What you get is a single database mounted by as many instances as cluster nodes you have configured. So every time your application establishes a db connection, it may be served by either cluster node. The data files can be in a clustered file system (like OCFS), in raw partitions or even in NFS mounts. There are a few guidelines to develop apps meant to run in a clustered database: http://www.cise.ufl.edu/help/database/oracle-docs/rac.920/a96598/design. htm Ivan. -Original Message- From: Viral_Thakkar [mailto:[EMAIL PROTECTED] Sent: Friday, August 27, 2004 7:35 AM To: Struts Users Mailing List Subject: [OT] Clustering Hi All, I am developing a J2EE application using Oracle 9iAS. I am looking for clustering information. I am looking for info on database side. * Are Database servers clustered in an ACTIVE/ACTIVE or ACTIVE/PASIVE mode? * Is the Oracle database supposed to reside on a file system or on a RAW partition? TIA. Regards, Viral - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [OT] Clustering Application Scope Objects
Hi Mike, This is an interesting question. Using the PlugIn interface to store "semi-static" data in the app context is a commonly accepted pattern (we use it in the project I'm working on now). However, there's no mechanism currently available in Struts to allow for data model change event notification (evidently, that's the major difference between Struts & Barracuda, from what little I understand of Barracuda). I'd personally love it if there were a simple mechanism available to Struts for just such event notification. In the meantime, you may want to store such data in session scope, as it's more "volatile" (periodically speaking). I don't know the nature of your data, but if it isn't a huge amount, it shouldn't be a problem. Keep us posted on your progress. Curtis -- cee dot tee at verizon dot net Mike Duffy wrote: Thanks Christian, Let me give you a specific example of what I am trying to do. I have several different sets of label/value pairs that are stored in database tables. These label/value pairs make up the select options and other interface options for web pages in the application. The options do not change that often, so rather than going back to the database each time a page is displayed, I'd like to store the options in maps that are stored in application scope and easily accessible through JSTL. On the occasion that the options do change, I'd like to reload the applicable maps to all servers in the cluster. It seems like this would be a good feature for Tomcat to have, but from reading the documentation it seems that the feature is not there. Perhaps some sort of JMS broadcast service would do the trick. Does any one have any suggestions? Mike - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [OT] Clustering Application Scope Objects
Hi Mike, Your JMS approach sounds fairly sophisticated; could take a while to implement ;-) Hibernate provides some out-of-the-box cacheing implementations, some of which take on this problem. I think you'll find this portion of their user guide fairly relevant: http://www.hibernate.org/hib_docs/reference/html_single/#performance-s3 In particular, the SwarmCache and JBoss TreeCache look like they could do what you're looking for. I bet they could be used outside Hibernate too, but I'm not sure. Matt Mike Duffy wrote: Thanks Christian, Let me give you a specific example of what I am trying to do. I have several different sets of label/value pairs that are stored in database tables. These label/value pairs make up the select options and other interface options for web pages in the application. The options do not change that often, so rather than going back to the database each time a page is displayed, I'd like to store the options in maps that are stored in application scope and easily accessible through JSTL. On the occasion that the options do change, I'd like to reload the applicable maps to all servers in the cluster. It seems like this would be a good feature for Tomcat to have, but from reading the documentation it seems that the feature is not there. Perhaps some sort of JMS broadcast service would do the trick. Does any one have any suggestions? Mike --- Christian Bollmeyer <[EMAIL PROTECTED]> wrote: On Saturday 10 April 2004 21:48, Mike Duffy wrote: Hi, I've read documentation for The Tomcat 5 Servlet/JSP Container: Clustering/Session Replication HOW-TO http://jakarta.apache.org/tomcat/tomcat-5.0-doc/cluster-howto.html I understand clustering for individual user sessions. Are there any correlated methods for clustering application scope objects? AFAIK there's nothing in the Servlet API specifications in this direction. Sessions may be swapped in and out by several different means (serialization or database), but application scope information is kept individually for each JVM. The J2EE API for the Interface ServletContext states, "In the case of a web application marked "distributed" in its deployment descriptor, there will be one context instance for each virtual machine. In this situation, the context cannot be used as a location to share global information (because the information won't be truly global). Use an external resource like a database instead." Yes. But then, it makes no difference whether your application is marked as 'distributable' or not when considering application scope objects. Application scope is always limited to a single JVM. 'Distributed' is about sessions only. Rather than use a database, what I would like to be able to do is make a call to servlet.getServletContext().setAttribute(key, object); and have the object stored in the application scope of all servers in the cluster. That surely would be desirable, but I don't know of any 'official', declarative means to achieve such behavior. Then: automatically synchronizing application scopes between several JVM instances would cause a lot of additional synchronization troubles - who determines what's the actual system state at a given point, e.g. if you update an application scope object (that is globally shared) on one server, and another user does likewise on a second one, what should be the end result? And so on. I know that EJBs were designed to serve this purpose; however, I would like to bypass the overhead and complexities of EJBs. EJBs live in the backend and are independant from the web tier. Looking at the problem from a web tier view (which is what Struts is about), it should not make any difference whether the information comes directly from a database in the end or from an EJB tier in-between. That's transparent. Anyway, using EJBs won't solve your problem in the end, for it's an entirely web tier specific issue, and EJBs are not part of that. If there isn't a switch that can be flipped in Tomcat, there might be a way to create a lightweight JMS administration class to serve this purpose. Has anyone tried this? There is no switch (remember the spoon boy in Matrix I ? :-). But the JMS idea is interesting. Never actually tried it, but of course, you can always handle application specific logic programma- tically, including using JMS for updating application scopes in a distributed environment. But I never tried. If the answer to this question is RTFM, please send a link; I've looked through the documentation and I can't seem to find a clear reference. Thanks for your time and consideration. Just some thoughts that came to me on early Easter Sunday :-) Let's better say: night; it's 2:16 local time here. So I should better go to bed. Mike HTH, -- Chris. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-
Re: [OT] Clustering Application Scope Objects
Thanks Christian, Let me give you a specific example of what I am trying to do. I have several different sets of label/value pairs that are stored in database tables. These label/value pairs make up the select options and other interface options for web pages in the application. The options do not change that often, so rather than going back to the database each time a page is displayed, I'd like to store the options in maps that are stored in application scope and easily accessible through JSTL. On the occasion that the options do change, I'd like to reload the applicable maps to all servers in the cluster. It seems like this would be a good feature for Tomcat to have, but from reading the documentation it seems that the feature is not there. Perhaps some sort of JMS broadcast service would do the trick. Does any one have any suggestions? Mike --- Christian Bollmeyer <[EMAIL PROTECTED]> wrote: > On Saturday 10 April 2004 21:48, Mike Duffy wrote: > > Hi, > > > I've read documentation for The Tomcat 5 Servlet/JSP Container: > > Clustering/Session Replication HOW-TO > > http://jakarta.apache.org/tomcat/tomcat-5.0-doc/cluster-howto.html > > > > I understand clustering for individual user sessions. Are there any > > correlated methods for clustering application scope objects? > > AFAIK there's nothing in the Servlet API specifications in this > direction. Sessions may be swapped in and out by several > different means (serialization or database), but application > scope information is kept individually for each JVM. > > > The J2EE API for the Interface ServletContext states, "In the case of > > a web application marked "distributed" in its deployment descriptor, > > there will be one context instance for each virtual machine. In this > > situation, the context cannot be used as a location to share global > > information (because the information won't be truly global). Use an > > external resource like a database instead." > > Yes. But then, it makes no difference whether your application > is marked as 'distributable' or not when considering application > scope objects. Application scope is always limited to a single > JVM. 'Distributed' is about sessions only. > > > Rather than use a database, what I would like to be able to do is > > make a call to > > > > servlet.getServletContext().setAttribute(key, object); > > > > and have the object stored in the application scope of all servers in > > the cluster. > > That surely would be desirable, but I don't know of any > 'official', declarative means to achieve such behavior. > Then: automatically synchronizing application scopes > between several JVM instances would cause a lot of > additional synchronization troubles - who determines > what's the actual system state at a given point, e.g. > if you update an application scope object (that is > globally shared) on one server, and another user > does likewise on a second one, what should be > the end result? And so on. > > > I know that EJBs were designed to serve this purpose; however, I > > would like to bypass the overhead and complexities of EJBs. > > EJBs live in the backend and are independant from > the web tier. Looking at the problem from a > web tier view (which is what Struts is about), > it should not make any difference whether the > information comes directly from a database > in the end or from an EJB tier in-between. > That's transparent. Anyway, using EJBs > won't solve your problem in the end, for > it's an entirely web tier specific issue, > and EJBs are not part of that. > > > If there isn't a switch that can be flipped in Tomcat, there might be > > a way to create a lightweight JMS administration class to serve this > > purpose. Has anyone tried this? > > There is no switch (remember the spoon > boy in Matrix I ? :-). But the JMS idea is > interesting. Never actually tried it, but > of course, you can always handle > application specific logic programma- > tically, including using JMS for updating > application scopes in a distributed > environment. But I never tried. > > > If the answer to this question is RTFM, please send a link; I've > > looked through the documentation and I can't seem to find a clear > > reference. > > > > Thanks for your time and consideration. > > Just some thoughts that came to me > on early Easter Sunday :-) Let's better > say: night; it's 2:16 local time here. So > I should better go to bed. > > > Mike > > HTH, > -- Chris. > > - > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > __ Do you Yahoo!? Yahoo! Small Business $15K Web Design Giveaway http://promotions.yahoo.com/design_giveaway/ - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [OT] Clustering Application Scope Objects
On Saturday 10 April 2004 21:48, Mike Duffy wrote: Hi, > I've read documentation for The Tomcat 5 Servlet/JSP Container: > Clustering/Session Replication HOW-TO > http://jakarta.apache.org/tomcat/tomcat-5.0-doc/cluster-howto.html > > I understand clustering for individual user sessions. Are there any > correlated methods for clustering application scope objects? AFAIK there's nothing in the Servlet API specifications in this direction. Sessions may be swapped in and out by several different means (serialization or database), but application scope information is kept individually for each JVM. > The J2EE API for the Interface ServletContext states, "In the case of > a web application marked "distributed" in its deployment descriptor, > there will be one context instance for each virtual machine. In this > situation, the context cannot be used as a location to share global > information (because the information won't be truly global). Use an > external resource like a database instead." Yes. But then, it makes no difference whether your application is marked as 'distributable' or not when considering application scope objects. Application scope is always limited to a single JVM. 'Distributed' is about sessions only. > Rather than use a database, what I would like to be able to do is > make a call to > > servlet.getServletContext().setAttribute(key, object); > > and have the object stored in the application scope of all servers in > the cluster. That surely would be desirable, but I don't know of any 'official', declarative means to achieve such behavior. Then: automatically synchronizing application scopes between several JVM instances would cause a lot of additional synchronization troubles - who determines what's the actual system state at a given point, e.g. if you update an application scope object (that is globally shared) on one server, and another user does likewise on a second one, what should be the end result? And so on. > I know that EJBs were designed to serve this purpose; however, I > would like to bypass the overhead and complexities of EJBs. EJBs live in the backend and are independant from the web tier. Looking at the problem from a web tier view (which is what Struts is about), it should not make any difference whether the information comes directly from a database in the end or from an EJB tier in-between. That's transparent. Anyway, using EJBs won't solve your problem in the end, for it's an entirely web tier specific issue, and EJBs are not part of that. > If there isn't a switch that can be flipped in Tomcat, there might be > a way to create a lightweight JMS administration class to serve this > purpose. Has anyone tried this? There is no switch (remember the spoon boy in Matrix I ? :-). But the JMS idea is interesting. Never actually tried it, but of course, you can always handle application specific logic programma- tically, including using JMS for updating application scopes in a distributed environment. But I never tried. > If the answer to this question is RTFM, please send a link; I've > looked through the documentation and I can't seem to find a clear > reference. > > Thanks for your time and consideration. Just some thoughts that came to me on early Easter Sunday :-) Let's better say: night; it's 2:16 local time here. So I should better go to bed. > Mike HTH, -- Chris. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
