Artima has a feature entry on Project Shoal which just reached 1.0 Early Access. Project Shoal is a Group Management System heavily used in GlassFish v2 but it can also be used in other contexts. The JXTA implementation used by default and which brings close-to-zero clustering configuration in GlassFish can also be replaced (by JGroups for instance) using the appropriate SPIs.

One of the reasons for the GlassFish v2 delay was the time needed to make sure we could go through as many customer scenarios as possible, specifically in the clustering area. Shreedhar reports on the 8-node GlassFish clustering quality tests which led to improved quality of Shoal itself. This and future use of Shoal in SailFin (the SIP supported application server based on GlassFish with Ericsson as a major contributor) shows the greater GlassFish Community at work.

Alexis has put together a nice summary of the basic concepts needed to use GlassFish v2 in a cluster profile. These include the concepts of domain, cluster, DAS and NA, master password, upgrading profiles, difference between cluster and enterprise profiles and more. Check all of these at Alexis' blog. And here you can find the transcript to Young Frankenstein (the young master).

JMX is the popular management and monitoring technology which has been integrated into Java SE starting with version 5. JDMK (Java Dynamic Management Kit) is a Sun product building on JMX to provide a distributed model with Federated MBean Servers, Agent Discovery, and SNMP support. This is a mature product that reached version 5.1 patch level 3 and it is now being open sourced under the name OpenDMK. Here's a good blog on all things JMX and OpenDMK.

Sreeni reports how the open sourcing of JDMK impacts GlassFish. The clustering architecture of version 2 uses cascading of instance MBeans on to the DAS (Domain Admin Server) and JMX remote connectivity. The latest GlassFish v2 build now uses the OpenDMK bits instead of the JDMK jars. This is transparent to the user and probably the last step on the way to a full open source GlassFish application server.

Think that enterprise features like clustering, fail-over, and load-balancing are beyond the scope of an open source application server? Think again. GlassFish v2 implements them all, and Nazrul has put together a nice writeup of these new features in his blog. GlassFish v2 is currently in Beta, and we're getting pretty close to its production release target (September). So it's a great time to try these new features out for yourself. (And if you need even more motivation, don't forget that someone is going to win a great prize just for sharing their experiences in trying out GlassFish.)

Kedar has a very good write-up on Clustering in GlassFish v2 that describes the basic use-case for a cluster profile and then shows how to upgrade to it from the usual developer profile. Kedar also has a two-parter screencast based on the writeup. The first part is only audio but the second part has the actual upgrade screens.

GlassFish is moving farther in the deployment level, with GlassFish v2's support for clustering and other enterprise-level features. A recent example of this is described by Prashanth where, triggered by questions at the JavaOne pavillion, he describes how to use a hardware appliance (Big-IP from F5 Networks) for load balancing. Prashanth's writeup is very detailed and explains how to configure Big-IP and also includes links to relevant GlassFish documents. Check it out!

Once upon a time... Open Source application servers had poor administration consoles, but GlassFish v2 has an excellent one; thanks to Anissa (on the right)! Anissa has started describing the Admin GUI; check these two blogs. The first came just before JavaOne and showed how to create a cluster on two machines (blog, screencast). In the second she describes a common scenario: upgrading from a development profile to a cluster profile (blog, screencast).

Great job! Check it out!

Say you want to debug a clustered deployment. So you open your laptop, start your appserver cluster with debugging enabled, then start one instance first, then the second instance, and ... wham! you have a port collision because both instances try to open the same jdwp port! But this has changed... Byron reports that starting in b45 (downloads) this has been fixed: you can start the instances and delay the debugging port binding. The syntax is not (yet?) pretty, but it is very useable. Check out Byron's blog for the details!

BTW, I see that the last posted build is GFv2b 43; I'll check the status on b45 and will update this blog.

Its nice to see more postings, as blogs or in aliases like the USERS mailing list, for the enterprise features in GlassFish v2. Kedar just posted a simple Recipe for Creating Clusters. We will be covering the enterprise features in GlassFish v2 during GlassFish Day and at JavaOne. Come join us; both events will be fun, and GlassFish Day is even free (but requires advanced registration).

Daniel has a very Complete Post on clustering of Web Apps on GlassFish. He covers installing GF, setting up the domains, setting up the cluster and deploying the app. To really take advantage of clustering you need to put a load balancer in front, something like Sun Web Server LB or Apache with mod_jk. Dan promises more blogs. It is nice to see more examples and documentation on clustering and failover on GFv2. The beta2 release (due by JavaOne) will be very functional and we will have tested on "small" clusters (4-8); expect more testing and in larger clusters before the final release. All performance numbers so far are very nice.

Sequoia is a follow-up to C-JDBC project (under LGPL) and now uses ASL) provides transparent database clustering solution that addresses load balancing and failover for any database.

Sequoia works with a number of containers, and that now includes GlassFish! Jagadish latest writeup includes detailed steps for how to successfully configure NetBeans, Sequoia, GlassFish and Derby. For details, check out Jagadish's blog. Enjoy!

The code for Shoal is now available. Shoal is a generic Dynamic Clustering framework (Overview) that is being used in GlassFish V2 but, like Grizzly and others, can be used independently. Shoal includes a client API and an SPI to plug-in different implementations. The project has been a collaboration with the JXTA team and the default implementation takes very good advantage of the features of JXTA to provide a highly-decentralized and highly scalable implementation.

Read Shreedhar's intro to the project and Mohamed's detailed description of its use of JXTA. In separate entries, Max and Bernard provide context and background on the project. An excellent addition to the GlassFish Community.

Shoal will be integrated in the next GlassFish v2 milestone, together with Larry's in-Memory Replication machinery.

Terracotta provides clustering and caching for Java objects by instrumenting the Java byte-codes and doing things like (pre)fetching content or updating copies. The company seems to be in an promotional tour and, according to this JDJ article, they are specifically mentioning GlassFish as a container.

Terracotta seems an interesting technology; if you have experiences with it or similar technologies, pray share them. It would be informative to do an experiment similar to what they did with Tomcat. Their web site has plenty of good material, including a couple of screencasts. Check it out!

Profile support is a key mechanism to make GlassFish V2 usable to simple developers and complex deployments. Kedar will be presenting the current proposal today, Wednesday, at 9am Pacific (GMT-7) time. I meant to post this yesterday but I scheduled it for the wrong day, my apologies for the short notice! Check the One Pager, and participate in the concall. For more information and future meetings, check the UE Wiki page.

Now, here is a question for the community. I would like to see more Non-Sun participation in these discussions, and I'd welcome suggestions on how to improve that. The current approach used by the GlassFish team relies on concalls to quickly cover the material initially, but a concall requires synchronicity and makes it hard for a casual or latecomer to participate. Groups like Apache rely on mailing lists, but these can drag over long periods of time - and email conversations do not work very well over +11 TZs.

In the brave new world of Wikis, forums, flashd demos, etc... do you have suggestions on how to increase participation without substantially slowing down development? Let us know your ideas so we can start trying them out.

The second Milestone for GlassFish V2 is available. This release is part of our Continuous Improvements approach; you can check the Highlights for M2 and then Download it. We have updated the QuickStart guide, as well as the Architecture Document. You can also checkk the full GF V2 Roadmap This release improves substantially on Milestone 1 and many bugs were fixed. It adds MacOS X support for clustering and integrates a first batch of WSIT (ProjectTango) features. The download size is also smaller thanks to Pack200.

I will double-check that the highlights are complete and will add any missing to the Wiki page. We are still getting used to using CVS-News and we don't use that all the time.