Piotr Woloszyn

Close to two years ago I had my first look at Terracotta when researching available technology for sharing state and scaling Aloha. First impressions were good. Terracotta delivered what promised: easy integration with an existing single node application and virtual shared heap available across whole cluster. However it has much more to offer. I have been looking again recently in the distributed cache and data grids area trying to assess options for a project I am working on currently. Below is a digest from my web-crawling of publicly available data on the topic. In each of the paragraphs I will try to put some light on a specific aspect that someone would like to know when considering Terracotta.

Resilience

Terracotta clusters offer simple and scalable HA implementations based on the Terracotta server array. Terracotta server arrays can vary from a basic two-node array consisting of an active server instance backed by a hot standby (for instant failover), to a multi-node array with many active server instances providing scale and deep failover coverage. Terracotta give high availability of the data by not only replicating it across servers for redundancy, but also by providing a simultaneous permanent storage of all current shared (in-memory) data to disk – available to all server instances (no loss of application state). Temporarily disconnected server instances and clients are automatically reconnected, hot standbys restored without operator intervention. There is no single point of failure and bottleneck in the Terracotta architecture.

Failover

Terracotta server array provides instant failover and continuous uptime and services using a hot standby or multiple active servers. Multiple Terracotta servers can be clustered together in active-active or active-passive configurations, helping to further scale Java applications and to provide disaster recovery.

DB Persistence

Terracotta helps to protect database in a scale out architecture. There is a problem with usual approach with a cache in front of the database. The performance of the cache is degradated when the number of nodes increases. Hibernate offers a second-level cache option, which can improve the performance of the Hibernate application by eliminating the need to fetch often used data objects from the database, instead storing, or caching, that data in process with your application. Without any mechanism to provide replication of the underlying cache, data written in one node can be stale on another node. This problem is not solved by Hibernate, rather it is up to the underlying cache to provide a replication scheme to keep multiple copies synchronized. Terracotta provides a consistent mechanism to keep any cache written in Java synchronized, and therefore provides this support for caches such as EHCache. Advantages: Performance — Reading and writing to replicated data with Terracotta can be 10x - 100x faster than other solutions, Cluster-wide load — Replication of data eliminates the need for more than one node to load data from the database. In a scaled-out application, this can reduce database load by a significant factor due to the fact that only one node needs to request any piece of data for a cache miss and all subsequent requests on other nodes will result in cache hits.

Distributed caching implementation

Terracotta uses concept of the Network-Attached Memory. It clusters JVMs directly underneath applications providing distributed, shared heap. The implementation doesn’t rely on serialization of whole objects, rather on instrumentation of the byte code and injecting hooks in there which allow synchronizing of deltas in the object state. State changes in a Client JVM are published to Terracotta Server Array and all other Client JVMs are seamlessly updated. Other Terracotta’s features apart from shared heap are: distributed locking, distributed data structures – (HashMap, ArrayList, LinkedBlockingQueue, any POJO), guaranteed coherent clustered objects. Objects do not have to implement Serializable to work with Terracotta.

OS/Platform restrictions

Terracotta is a 100% Java solution.

Certified platforms: Solaris 10, Solaris 9, RedHat ES4, SUSE ES10.1 (32/64bit), Windows Server 2003 R2, WindowsXP (dev only)
Server JDKs: Sun Hotspot 1.5.0_16, Sun Hotspot 1.6.0_07
Client JDKs: Sun Hotspot 1.4.2_18, Sun Hotspot 1.5.0_16, Sun Hotspot 1.6.0_07
Containers: Tomcat 6.0.18 / 5.5.26 / 5.0.28, Apache Geronimo 1.1.1, BEA WebLogic 8.1 sp6, BEA WebLogic 9.2.MP2, JBoss 3.2.8.sp1 / 4.0.5, Jetty 6.1.11, WebSphere CE 1.0.1.2 / AS 6.1.0.7, Glassfish V1 ur1-p01-b02, Glassfish-v2 ur2-b04

Ease of implementation

Big advantage of Terracotta. A clustered application looks no different than a standard Java applications. In theory a correct multi-threaded application can be clustered just by providing a simple xml config defining roots of application state hierarchy which one wants to share across multiple JVMs. Distributed locking and thread coordination uses usual language artefacts like ’synchronized’, ‘wait’/'notify’ and java.util.concurrent support. Pre-existing Terracotta Integration Modules (TIMs): purpose-built, tested, and tuned to implement a particular clustered architecture pattern or integrate with a particular third-party technology, for example use TIMs for Spring, Hibernate, Quartz, Compass, Lucene, Guice, Camel, Struts, Joda time and more to seamlessly add clustering to an application. Last but not least: Terracotta is an open source product with an active community.

Hardware requirements

Client:

Memory Sizing - Java Heap: each client JVM should do work on data that fits in its local heap
Memory Sizing - RAM: enough RAM to accommodate the heap of the Terracotta client JVM
Number of client nodes: determined by the total size of the active clustered data set and total required throughput. Entire data set will fit in the sum of the total heap available for clustered object storage across all Terracotta client JVMs.

Server:

Memory Sizing - Java Heap: Ideally, all of the clustered object data should fit into its heap, note: allow for garbaged objects, not recommended heap > 6/8GB
Memory Sizing - RAM: enough RAM to accommodate the heap of the Terracotta server JVM. In addition, there should be enough RAM headroom to keep the blocks of the Sleepycat data store files in the host operating system’s filesystem cache for fast access.
CPU: dedicated server, capacity planning needed. Note: The more garbage your application creates, the more CPU resources will be consumed by the Terracotta Distributed Garbage Collector. Likewise, the higher the object read or write rate between Terracotta client nodes and a Terracotta server instance, the more CPU resources that server instance consumes.
Number of server nodes: min 2 for HA and failover

Network

Gigabit full duplex between server and clients, and server array, low latency network, redundant NICs and redundant switches with automatic failover

Performance

According to Terracotta: it is faster than other distributed cache, data grid solutions. It is due to the nature of synchronizing only deltas of object state. Terracotta scales much better than i.e. Coherence even with big transaction payloads. At small scale it has throughput 2x Oracle’s at all transaction sizes. At high scale (8 nodes) throughput was as much as 10x higher than Oracle. However independent tests of a specific use case scenario should be performed to back up the results.

Support package

There are Enterprise Subscription packages: production support GOLD 7X24 | SILVER 5X12 with issue escalation privileges and procedures. Also consulting and training services. Pricing individually discussed.

Code stability

Current version 2.7.1. Code is mature as the company started to work on Terraccota in 2003. Two years ago project has been opensourced. Licensed as Terracotta Public License, based on the Mozilla Public License. Free version requires displaying the words “Powered by Terracotta” in user interfaces. Commercial Version removes this requirement. Typical price $7,000 per node. A single Terracotta server can handle up to 150 nodes, but on heavy Java applications, such as those used in online reservation systems, that gets reduced to somewhere between 15 to 20 nodes per Terracotta server.

Partitioning

Workload partitioning:

Terracotta does not ship with a particular workload partitioning strategy in the box. Layer 7 sticky load balancer that is able to route requests for a particular user session to the same Terracotta client application server and is likewise able to balance user sessions evenly across all available application servers can be used to achieve workload partitioning. Use of round-robin or a non-sticky load balancing is strongly discouraged. Other forms of workload partitioning may be required for different use cases. Implementing workload partitioning algorithms in Java that can then be made cluster-aware via Terracotta is often a simple way to meet custom workload partitioning requirements.

Data partitioning

Each application server should only have the data that its workload requires access to in heap. (Magic warning!) Terracotta detects how your application uses its data and optimizes the data location automatically to minimize network updates. Terracotta manages replicating in the most efficient way. Only deltas to your object graph are pushed, objects automatically reside in JVMs that need them, and not in JVMs that don’t (vs. alternatives that always force you to configure and partition up front).

Transactional

Terracotta transactions are sets of clustered object changes that must be applied atomically. Transactions are bounded by lock acquisition and release. When a clustered lock is acquired by a thread, a Terracotta transaction is started and all changes to clustered objects made within the scope of that lock are added to that transaction. When the lock is released, the Terracotta transaction is committed. If a thread attempts to modify a clustered object outside the context of a terracotta transaction, a runtime exception will be thrown.

Replication and consistency

Fully synchronous replication between nodes. The data is always coherent and doesn’t suffer any cache drifts.

Popularity: 64% [?]

Share This

Me and my new team preparing now to starting a new greenfield project in month or so. Starting something new is very exciting itself and lets us put new processes in place. But how to decide about the processes? Simply by collaborating, discussing and exchanging ideas and previous experience. In this post I will focus on the development/release process and choices we are facing at the moment.

Just to give some perspective… In the past when working on the BT SDK with Web21c team I was using ScrumWorks for managing user stories or, as some other people would call it, features, improvements or bugfixes and mapping those to particular releases. The public releases were very infrequent and scheduled every 3 months. Emergency releases with bug fixes could happen more often. Of course as for an agile team internal releases occurred more often, usually every 2 weeks ended with an acceptance session with a product owner. The Subversion has been used as the source repository, with all active development happening usually in parallel in the trunk. Some time before release the code was frozen in a branch by copying an active trunk and only bug fixes were applied to the release branch, subsequently merged back to the trunk. I could call this process “Develop in trunk, release from a branch”.

At my current project we use only JIRA for managing bugs/features/improvements and matching it to particular releases and the releases will occur much more often (every week or two) with possible multiple emergency releases in meantime. Source is kept in Subversion as well. The team is not calling itself “agile” however I think it has to be very agile to cope with the schedule and a right process should be established to support it. It could look like:

Development time

1. In JIRA we create a bunch of tickets for new features, bug fixes or improvements. A developer is assigned (hey, who likes that phrase?) a task to work on.
2. Before starting work a new branch is created from trunk with the JIRA number as a name
3. Work on that task commences and all changes are committed to the new branch. If there are several dependant tickets then JIRA should have a parent ticket created and all dependant tickets assigned as children. It resembles higher level user story with tasks from Scrum.

Build/Release time

1. In JIRA a release or version will be set up and given a name (i.e. 1.0.1). This will contain a list of tickets that are to be used for this release
2. A new branch will be created in SVN from trunk, matching the name of the JIRA release.
3. The branches that match the JIRA tickets for this release are merged into the new release branch.
4. A continuous integration build is pointed to the newly created release branch.
5. The build should pass unit tests, integration tests, automated acceptance tests, is then deployed to UAT environment for final user acceptance tests.
6. When ready for release, a tag is created from the branch matching the JIRA release name. The software is deployed to production.
7. The code from the tag is now merged back to trunk

It is clearly “Develop in branches, release from trunk”. Sounds weird at the beginning. Have to admit that I didn’t use exactly this process before but my first impression is that it is really structured methodology which:

• gives superb view what actually is currently deployed (via mapping tickets to releases in JIRA)
• allows adding/removing changes easily – we can always take a previous tagged version and apply selected modification by merging
• supports external auditing

In the same time my gut feel tells me that:

• merging before release might be a nightmare (volunteers for a release manager?)
• the process seems to be too regulated, “boxed”, “walled”

It will very much depend on the release schedule and how often the changes will be released which means how often the trunk will be up to date with the latest code. Let’s assume we have biweekly releases. It will mean that at the average a developer will be working with a one week old code. With an application in a maintenance mode with occasional bug fixes which doesn’t change dramatically it would be acceptable but taking as example a new project with many contributors, dependencies between components and a lot of expected changes in design and implementation I think that might be problematic.

It really doesn’t encourage refactoring neither. Just imaging what would happen if two branches taken from the same version of TRUNK were exposed to wider and more intensive refactoring? And for me refactoring is a necessary factor which means continuous and gradual code quality improvement leading to system quality improvement. I know, we should do everything right first time. Why I didn’t think about this before?

Other downside which I can see is that we might completely lose transparency and visibility what is happening with the code. People will commit to “own” branches without a continuous integration. I have to say that I like to monitor changes to current source code via email from continuous integration with last check-in comments. It gives me a way of quickly assessing if the ongoing development is related to my current task, I can merge incoming changes one by one with my changes which means that I don’t do something which someone already has done or something which doesn’t make sense any more.

So, I have mixed feelings. The biggest advantage of the process I can see is easy adding/withdrawing pieces of functionality at short notice. And it (at least in theory, not mentioning the merging hell) fulfils the role very well. The process actually resembles the way how open source communities work. Each contributor branches from trunk and does development on his/her own without breaking anyone else code. When feature is ready then it is (after code review) merged back to trunk or release branch.

It appears that the problem here is caused by subversion not being the best at the merging job. So as we can see a good tool for merging is essential here to make the life easier. It looks like open source guys moved towards GIT and BAZAAR now with GIT getting a lot of traction as it has support for SVN as well.

The process described is not final and would look into refining it to better support given development/release practises. If you have suggestions, comments or you have been using similar process please free to contribute and share.

Popularity: 42% [?]

Share This

I came across a strange behaviour of Selenium and wanted to share my pain with someone really. But first couple of words of introduction. I use Selenium in my current webapp project which I am working on. The application uses quite a lot of javascript (Dojo framework mostly) which can be tested using D.O.H. framework but the Selenium is giving us the final sanity checks, acts as integration test framework and may be used by product owner in an agile environment to drive acceptance on user stories via automated acceptance tests.

Selenium is an excellent package. I really love it. Just to mention integration with various browsers, Selenium IDE, support for integration with build (Selenese ant task) and really easy way of creating and storing the tests as HTML. It is a framework which cannot be ignored by anyone working on webinterfaces.

There is previously mentioned selenese ant task which helps with integration of the Selenium tests and an ant build. You can grab the jar from the Selenium RC distribution.

However as most of the things around us it is not a perfect software. Anyone had the famous “Permission denied to get property Location.href”? Cross domain scripting issues?
In my case I figure out two reasons for above problem and I wanted to share it:

Problem 1:

Let’s look at the following selenium command:

• open
• http://host/some/url

Looks good, doesn’t it? So it might work, but it can as well break on another box with the “Permission denied to get property” and some other scary warnings. The solution in this case was simply to remove the host definition from url leaving just:

• open
• /some/url

Problem 2:

One of the pages I was testing had redirection in case of the authorization violation. So esentially after single HTTP request the browser was receiving single response with redirection and then performed another request to the URL defined in the redirection. Selenium couldn’t handle that correctly. But again, it wasn’t deterministic, it could work on one box but fail on another. Timing issues etc. Hate that. Neither of the combinations worked in consistent way: openAndWait, open + pause.

The solution in this case required change in the webapp to remove usage of the redirection (which I believe was an ugly way of dealing with the problem anyway).

Hope it will help someone with similar issues. If you know another tricks or something to avoid when using Selenium please contact me or leave a comment here.

Popularity: 94% [?]

Share This

I used Google Trends to check which Linux distribution is the most popular among Google search users. As you could probably easily guess and what is visible on the diagram Ubuntu is clearly the winner at the moment with the biggest number of search requests. What is interesting is that at the beginning of 2003 there were at least three equally popular distributions: Debian, Suse (now OpenSuse) and Mandrake (now Mandriva) with Red Hat slightly behind. Situation changed since then though. Debian sort of maintains its share. But other distributions became less and less popular with Mandrake loosing the most but partially probably due to changed name. Ubuntu with its excellent online community gains attention with linear search volume increase. I hope it will remain like that and then it is a matter of time when Windows with its ridiculous Vista release will go to hell.

Popularity: 28% [?]

Share This

There are situations when you can’t put all data which you need in a single instance of a relational database. The reasons may differ. Maybe because it is too much of the data itself. Or there is a problem with network latency of a distributed architecture. Scaling? Or one of many other reasons. The answer is: horizontal partitioning. The process of splitting up your data sets that once has been called partitioning, now, it has a new name: sharding. If your data doesn’t fit on one machine you split it up into segments and each segment is called a shard. It is term used initially at Google but now spreading everywhere.

So… There is a new project Hibernate Shards which is a framework that is designed to encapsulate and minimize complexity of accessing multiple databases by adding support for horizontal partitioning on top of Hibernate Core. It was once the 20 percent project at Google but it is open-sourced now and licensed under the LGPL. If you know the Core Hibernate API you know the Shards API as its implementation hasn’t violated the Core API. Basic assumptions and paradigms for using Hibernate are still valid as Configuration, SessionFactory, and Session objects are almost exactly the same. Interfaces from Hibernate Core:

• org.hibernate.Session
• org.hibernate.SessionFactory
• org.hibernate.Criteria
• org.hibernate.Query

have shard-aware extensions:

• org.hibernate.shards.session.ShardedSession
• org.hibernate.shards.ShardedSessionFactory
• org.hibernate.shards.criteria.ShardedCriteria
• org.hibernate.shards.query.ShardedQuery

The implementations for these four interfaces serve as a sharding engine that knows how to apply an application-specific sharding logic. This logic is a set of rules how data is distributed across its shards. To specify this logic we have to implement the interfaces below:

• org.hibernate.shards.strategy.selection.ShardSelectionStrategy
• org.hibernate.shards.strategy.resolution.ShardResolutionStrategy
• org.hibernate.shards.strategy.access.ShardAccessStrategy

Hibernate Shards comes with a couple of simple implementations of these interfaces. For instance for the Shard Selection Strategy we have choice of sequential (load balanced round robin) or parallel access.

To finish the topic of the sharding logic I shall mention the id generation as well. As the standard database sequences can’t be used in distributed environment we have a choice of two primary key generators:

• ShardedUUIDGenerator - that generates a big random number
• ShardedTableHiLoGenerator - that uses a table in one of the shards to generate the primary key. Obviously it is a single point of failure for our system.

Although project is still in its early stage and even the creators warn about possible glitches this nice wrapper for hiding all the complexity of distributing data around multiple relational databases is very much worth looking at, especially if the Hibernate is already used as the ORM tool of choice.

Popularity: 43% [?]

Share This

I can recall a conversation with psd from early this year about why I think current way of uploading photos to flickr sux big time. I treat flickr as repository and backup of _all_ my photos (12525 photos at the moment, still growing) and not just as a photo blog. It means that very often I have to upload quite a lot of them in batches. The existing solutions just make me angry (how shabby is the standard Flickr Uploadr!) when I have to tag them and correctly name or create sets. I am not (yet) a mac user. Maybe there is something more user friendly, I hope.

So… Coming back to the conversation with psd I have told him that unless there is some clever integration with OS which will make the process of uploading more smooth it will be always the pain. I expected as well a bit more metadata to be populated by camera automatically. How cool would be to have GPS and geo-tag all the photos when taking them. That’s not the end. I would expect as well to be able to define some tags in a camera.

That would solve a bit problem with tags. The other problems are setting permissions, creating sets etc. For pretty long time I couldn’t find any tool which I liked. Lately after migrating all my home computers to Gutsy I found something called Flickrfs as an available package. It is a virtual filesystem which mounts a flickr account as any other data file storage. It synchronizes the flickr account with local filesystem and shows photos as images files with all metadata represented as text files.

Imagine that. You can upload photos to flickr just by using cp command, the same applies to downloading photos (even if Flickr itself makes it as hard as possible to get back of your own photos). Deleting is as easy as invoking rm. You can set permissions using standard chmod command and define sets by creating symlinks with ln.

I like the idea itself very much. The Flickrfs is created and maintained mainly by one person so there might be some small “bugletes” but I still support this project and wish it the best. Well done Manish Rai Jain!

Popularity: 25% [?]

Share This

Some of you who use Eclipse 3.3 with Sun JVM might notice that it crashes from time to time. Those who have many plugins certainly experienced this more often. Problem and solution for this bug is very trivial but worth mentioning: leaking permgen space. Comes back like November flu! Just to remind the permgen space is a memory for storing “data needed by the virtual machine to describe objects that do not have an equivalence at the Java language level. For example objects describing classes and methods are stored in the permanent generation“. More information about permgen can be found in an article about tuning garbage collection.

To fix the problem edit your eclipse.ini and add following lines:

-XX:PermSize=128M
-XX:MaxPermSize=256M

You might also try Eclipse 3.3.1.1 which includes a fix for the above problem.

Enjoy!

Popularity: 38% [?]

Share This

This blog entry will introduce you how to integrate AStyle into your Ant build.

Short introduction: In my current project I work on software which generates SDKs for multiple languages consuming Web21C services . The SDKs are SOAP webservice clients for multiple languages as Java, Python, PHP or .Net (new are comming soon). XSLT transformations are extensively in use to achieve the goal of generating those SDKs.

The problem is that we are facing a trade off:

• we have ugly formatted XSL to be able to generate correctly formatted output code
• we can have pretty formatted and easy to maintain and read XSL but not correctly formatted output in particular language

Let’s assume that we are not crazy yet and we will not follow the first option. The second option looks better but is still not perfect. We need to plug a tool which will do the dirty job for us: correct indentation, format and beautify our output source. Artistic Style (AStyle) has been selected as the tool as it is small, simple, fast tool and supports both Java and C#.

However the AStyle has couple of downsides:

• doesn’t have easy integration with Ant
• accepts only list of files for transformation, there is no recursive mode

Let’s assume that all source files to be transformed are in folder generated, with multiple subfolders. As there is no integration with Ant we will have to use <exec/> task. Something like:

<property name="astyle.exe" location="${astyle.home}/astyle.exe"></property>
<exec executable="${astyle.exe}" dir="${basedir}">
    <arg value="--mode=java"></arg>
    <arg value="--style=java"></arg>
    <arg value="--suffix=none"></arg>
    <arg value="${sourcefile}"></arg>
</exec>

AStyle accepts following arguments:

astyle [options] SourceFile1.java SourceFile2.java SourceFile3.java [...]

Wouldn’t it be nice to have -R option?

Our first approach could be construction of a space separated list of paths to all files in folder generated and then using this as argument in exec task. We would quickly realize that we have a possible problem with maximum length of a command line executed via exec task which is just 256 characters. Better will be to execute astyle many times, for each file to be formatted. Core Ant task are not enough in this case. We have to use ant-contrib tasks. First copy ant-contrib jar to your ANT_HOME/lib and add in your build:

<taskdef resource="net/sf/antcontrib/antlib.xml" />

It will let you use <for/> task which should help us to solve our problem as it will allow us to iterate over list of paths and execute external command for each one. The final ant task could look like:

<target name="format">
    <for param="file">
        <path>
  	    <fileset dir="${generated}" includes="**/*.java"/>
  	</path>
  	<sequential>
  	    <property name="astyle.exe" location="${astyle.home}/astyle.exe" />
	    <exec executable="${astyle.exe}" dir="${basedir}">
	        <arg value="--mode=java"/>
	        <arg value="--style=java"/>
	        <arg value="--suffix=none"/>
	        <arg value="@{file}"/>
	    </exec>
  	</sequential>
    </for>
</target>

Fist we define a name of the parameter to pass the tokens or files to the sequential, let’s call it just file. Note that for makes use of ant’s macrodef task, so the @{} notation is used for parameter substitution. Next we define path including all java files in generated folder. The last step is to use <sequential/> block where the execution of astyle.exe will happen for each source file defined using @{file} parameter.

Popularity: 40% [?]

Share This

Again about the Mylyn plugin. If you install Eclipse 3.3 Europa there isn’t, due to the Mylar -> Mylyn renaming, the web connector plugin which could be found in the Mylyn 1.0. This plugin is useful when we are working with an unsupported web based bug repository. The plugin still exists but can be found in a separate update site. Simply add new remote update site: http://download.eclipse.org/tools/mylyn/update/extras to find a set of Mylyn plugins including the web connector.

If I find some time I will try to connect to Scrum Works repository using that plugin. If someone has done it already please let me know.

Popularity: 33% [?]

Share This

I migrated last week to Eclipse 3.3.0 and during usual routine of installing plugins (must to have: JBossIDE, SpringIDE, Subclipse) I decided to try two plugins I haven’t been using yet. First is a subproject from Eclipse Tools project called Implementors. It fills a gap in standard Eclipse source code navigation which doesn’t really well support working with interfaces. With Implementors plugin it is possible to jump to implementation of an interface and to an interface of implementation. Sounds unimpressive yeah? But it is really useful plugin and I could call it a “Plugin of the Month in Category Simple but Cool”. If I had that category. So, whenever you edit a class which has dependencies defined as interfaces just hit Alt-F3 on a method from an interface to go to implementation. If you have more than one implementation you will have a choice. Ctrl-Alt-F3 will take you to interface. Thanks Andy for introducing that plugin!

The second plugin I wanted to write about is Mylyn formerly known as Mylar. I have heard about this plugin long time ago, talked about it with Tomaz many times but never actually tried it out. My bad and now I regret that. Because I found Mylyn just awesome. In couple of words Mylyn is a task focused ui for Eclipse which filters all unneeded information noise, leaving you only relevant information. It monitors your activity when working on a particular task and creates a context of that task which makes it is very easy to switch between tasks. No more navigating, searching, scrolling again. Mylyn integrates with repositories like JIRA, Bugzilla or Trac. There is also option of using local task repository.

As I use Scrum Works for task management I would like to use that repository with Mylyn but couldn’t find a connector for that one. Anyone solved that problem?

Popularity: 30% [?]

Share This