Thursday, December 17, 2009

3 Spring Stuff Worth Sharing

Firstly, Spring 3.0.0 is released not more than 2 days ago. Juergen (if you have the habit of reading source code, you shall know him) blogged about the features at Spring Framework 3.0 goes GA. I am particularly interested to explore more (possibly make changes to my existing project) in the area of @Configuration, SpEL, REST, OXM and validation.

I also upgraded my project this morning, with less than 10 lines of code changes to switch the plug from 2.5.6 to 3.0.0 - the compilation errors were some casting issues of generics type - yes, Spring 3 is now compile-time type safe. If you use Maven, you may want to read about the pom and repository configurations here.

Second-and-thirdly, below are some notes I wrote down in company's intranet. :)

Spring Resolves Dependencies and Initializes Accordingly, Not You

Before that, I thought implementing the Ordered interface could impact the order of which beans shall first be loaded. Then I realized, it doesn't work. Ordered is used.. as far as I can tell right now, two on top of my head: during invocation of Bean and BeanFactory post processing and the order of Advices - just not Bean initialization!

So how does it really work (for which bean to be loaded before any other)? One thing that I found (or maybe really just one): dependencies - bean references and @DependsOn/depends-on. When the BeanFactory preinstantiates the beans, it resolves its dependencies (via BeanDefinitionValueResolver) and eventually the beans who are referenced/dependent by others will be initialized first.

Do take note that, you rarely use @DependsOn unless you have static (*ugh!!*) references. For my case, HibernateSessionFactoryBean depends on SubEntityManager and those enums in it are accessed via the static (*ugh!!*) method SubEntityManager.getInstance() by the SubEntityUserType.

Spring Shutdown Hook and Unit Testing

A Spring ApplicationContext has a close() method, things like HibernateSessionFactoryBean will only be destroyed when the context is destroyed via the close() method - which is triggered by shutdown hook. See AbstractApplicationContext.registerShutdownHook(). So what is important here to know is, if you are running unit tests, a series of them, destroy can only happen when the shutdown hook is called.

Here's an example, I was running two test fixtures but the first one didn't shutdown until the whole test execution exited.
[12-16 16:26:45] INFO  GenericApplicationContext [Thread-4]: Closing org.springframework.context.support.GenericApplicationContext@1923ca5: ...startup date [Wed Dec 16 16:26:42 MYT 2009]; root of context hierarchy
[12-16 16:26:45] INFO  GenericApplicationContext [Thread-3]: Closing org.springframework.context.support.GenericApplicationContext@19b5217: ...startup date [Wed Dec 16 16:26:34 MYT 2009]; root of context hierarchy
[12-16 16:26:45] INFO  HibernateSessionFactoryBean [Thread-4]: Closing Hibernate SessionFactory
[12-16 16:26:45] INFO  HibernateSessionFactoryBean [Thread-3]: Closing Hibernate SessionFactory

Saturday, November 7, 2009

Wireless Network disappeared after Ubuntu 9.10 (Karmic) upgrade

Upgraded to 9.10 this morning but funny thing was - wireless networking wasn't working anymore - ath_pci module is gone from the release.

Fixed by loading the ath5k module, as suggested by the ThinkPad wiki. You can append it to /etc/modules for it to be automatically loaded.

Thursday, July 23, 2009

Measuring Your Engineers

To say that I am managing an Agile team is not entirely accurate, we are just trying to be agile. Anyway, skipping the story, the team's velocity increased more than 800% for the last 2 iterations, mainly due to management pressure. Put aside other reasons such as requirement gathering, design, headcount, etc.

1. Engineers are most probably working within their comfort zone.

For the past 3 months, I had passed/failed/extended the probation period of the engineers. During the session, they discussed with us their strength, weaknesses and areas for improvement.

2. Most engineers highlighted that they're lacking some knowledge to perform better.

So, last weekend, after having a discussion with the management, I was given a task - create KPIs to give engineers a direction as well as to provide management a way to measure their performance. So there are two things - KPIs and Performance Review. The first thing that got into my mind was - how did my ex-company do it?

In the support team that I worked in, we used a ladder-based system where there were 4 skill levels defined with skills to achieve. We would also set KPIs such as "resolving 30 support cases with customer satisfaction above 75%" or "to be SCJP certified". The progress, results and data would be presented to the management during our performance review. Trust me, people seldom stepped out the room wyith a happy face because the management loved to tell you where you did not do good enough.

I was thinking of the ladder since then.

I then looked into some blogs and Stack Overflow questions, such as this (Joel), this, this (Steve Yegge), this and this, which in general send the same message:

3. Management focuses too much on measuring the performance than improving the products and getting things done.

Something quite similar to this will be - timesheet and the reason I abolished it because it is irrelevant with the team velocity and business values. What does it mean then, shall we not measure the engineers at all? Your HR is certainly expecting some kind of metrics for her as a parameter to make salary adjustment.

What shall we look into then? Let's start from things that we shall never use to measure an engineer:
  • SLOC - Source Lines of Code, the dumbest ever metric. Good programmers write lesser lines.
  • Number of bug fixes - No way! The engineers would have figured out how to compromise this by writing more buggy code. I strongly against this just like how I disagree with giving points to bug fixes.
  • Hours logged - Timesheet is a waste of time.

I had like a 10-minute talk with the engineers yesterday to find out how they want to be evaluated and paid. I think this is important instead of me making decision and guesses based on how other software houses do it. It turned out that they want to be reviewed by the peers, by stories delivered, by code quality, and by having salary within different range they expect to be put an expectation on the skill set they possess.

So here is my summary, the performance metrics will be generally based on the expectations, the results, and the feedbacks:
  • The rating from their team members.
  • The rating from their direct supervisor, basically this person also knows the agility, code quality, velocity, etc. of this person.
  • The total stories delivered.
  • The skill set, which KPIs can be related with this, not entirely though.
Are KPIs necessary? KPIs shall not be mandated, they have to be discussed and proposed by the individuals, based on their comfort level and see how far we want to go above it. If a KPI is not going to help an engineer to produce quality code and better product, drop it. Also, if one doesn't want to play this KPI game as he really has a clear idea about every thing he is doing, don't force him to change the way he handles and manages his job and directions.

The last reminder is that, we shall not expect to have a perfect and fair formula that objectively measure and rate an engineer. Whether someone will be promoted or given a good salary shall be based on human judgments, as we are dealing with humans and humans make important decisions like you always do.

Friday, July 10, 2009

Hibernate session issue during Spring tests

I reckon this is something good to share as it took me a few hours to get the problem resolved.

We are using SpringJUnit4ClassRunner for our DAO tests, the tests passed when they were run in Eclipse, but not in Maven. One of the test classes failed with HibernateSystemException:

org.springframework.orm.hibernate3.HibernateSystemException: Illegal attempt to associate a collection with two open sessions; nested exception is org.hibernate.HibernateException: Illegal attempt to associate a collection with two open sessions

Google search gave me some clue about the problem had something to do with cascading style, transaction, and duplication but they didn't lead me anywhere. I broke at AbstractPersistentCollection.setCurrentSession to find out that the session/session factory (DAOs are instances of HibernateDaoSupport) assigned to one of the DAOs was different.

A second look at the test class which had:
@Autowired
BillingInvoiceDao $;
BillingOrderDao orderDao;
BillingItemDao itemDao;
MembershipDao membershipDao;

@Override
protected void setUpDataAccessObjects() {
membershipDao = DataAccessObjectHelper.getDataAccessObject(MembershipDao.class);
itemDao = DataAccessObjectHelper.getDataAccessObject(BillingItemDao.class);
orderDao = DataAccessObjectHelper.getDataAccessObject(BillingOrderDao.class);
}

I found out that BillingInvoiceDao was not a proxy instance but the other DAOs are. The log may explain something:
[07-11 09:19:30] DEBUG AutowiredAnnotationBeanPostProcessor [main]: Autowiring by type from bean name 'itest.net.aflexi.cdn.billing.HibernateBillingInvoiceDaoTest' to bean named 'billingInvoiceDao'
[07-11 09:19:30] DEBUG DefaultListableBeanFactory [main]: Returning cached instance of singleton bean 'org.springframework.transaction.config.internalTransactionAdvisor'
[07-11 09:19:30] DEBUG DefaultListableBeanFactory [main]: Returning cached instance of singleton bean 'org.springframework.transaction.config.internalTransactionAdvisor'
[07-11 09:19:30] DEBUG AnnotationTransactionAttributeSource [main]: Adding transactional method [testCrud] with attribute [PROPAGATION_REQUIRED,ISOLATION_DEFAULT]
[07-11 09:19:30] DEBUG AnnotationAwareAspectJAutoProxyCreator [main]: Creating implicit proxy for bean 'itest.net.aflexi.cdn.billing.HibernateBillingInvoiceDaoTest' with 0 common interceptors and 1 specific interceptors
[07-11 09:19:30] DEBUG JdkDynamicAopProxy [main]: Creating JDK dynamic proxy: target source is SingletonTargetSource for target object [itest.net.aflexi.cdn.billing.HibernateBillingInvoiceDaoTest@17afcff]
[07-11 09:19:30] DEBUG AnnotationTransactionAttributeSource [main]: Adding transactional method [testCrud] with attribute [PROPAGATION_REQUIRED,ISOLATION_DEFAULT]
[07-11 09:19:30] DEBUG TransactionalTestExecutionListener [main]: Explicit transaction definition [PROPAGATION_REQUIRED,ISOLATION_DEFAULT] found for test context [[TestContext@134b58c testClass = HibernateBillingInvoiceDaoTest, locations = array['classpath:spring/dataAccessTestContext.xml', 'classpath:spring/dataAccessTestContext.billing.xml'], testInstance = itest.net.aflexi.cdn.billing.HibernateBillingInvoiceDaoTest@17afcff, testMethod = testCrud@HibernateBillingInvoiceDaoTest, testException = [null]]]
[07-11 09:19:30] DEBUG TransactionalTestExecutionListener [main]: Retrieved @TransactionConfiguration [@org.springframework.test.context.transaction.TransactionConfiguration(defaultRollback=true, transactionManager=mainTxManager)] for test class [class itest.net.aflexi.cdn.billing.HibernateBillingInvoiceDaoTest]
[07-11 09:19:30] DEBUG TransactionalTestExecutionListener [main]: Retrieved TransactionConfigurationAttributes [[TransactionConfigurationAttributes@ddc524 transactionManagerName = 'mainTxManager', defaultRollback = true]] for class [class itest.net.aflexi.cdn.billing.HibernateBillingInvoiceDaoTest]
[07-11 09:19:30] DEBUG DefaultListableBeanFactory [main]: Returning cached instance of singleton bean 'mainTxManager'
[07-11 09:19:30] DEBUG TransactionalTestExecutionListener [main]: Executing @BeforeTransaction method [public void net.aflexi.cdn.core.test.AbstractDaoTest.setUpBeforeTransaction() throws java.lang.Exception] for test context [[TestContext@134b58c testClass = HibernateBillingInvoiceDaoTest, locations = array['classpath:spring/dataAccessTestContext.xml', 'classpath:spring/dataAccessTestContext.billing.xml'], testInstance = itest.net.aflexi.cdn.billing.HibernateBillingInvoiceDaoTest@17afcff, testMethod = testCrud@HibernateBillingInvoiceDaoTest, testException = [null]]]
[07-11 09:19:30] DEBUG DefaultListableBeanFactory [main]: Returning cached instance of singleton bean 'membershipDao'

That transaction propagation happened after autowiring of beans. The fix of my problem is to @Autowired all DAOs (so that they are consistent) or assign them via the DataAccessObjectHelper as shown in the snippet.

Hope this helps.

Saturday, April 11, 2009

"Autoboxing" Array is Illegal

Wrote some array methods today then I had a question about auto-converting an int[] to Integer[] (auto-boxing) while I was driving back home.

For instance, this works,
public static  V wrapper(V v, Class clazz){
return v;
}

int i = wrapper(1, Integer.class);

But this is not compilable,
public static  V[] wrapperArray(V[] v, Class clazz){
return v;
}

int[] i2 = wrapperArray(i2, int.class);

int.class will be auto-converted to Integer.class but int[] will not. I can't find the exact line in JLS describing this, do leave a comment if you could.

Friday, April 10, 2009

Strange Maven Plugin Metadata Problem

Not too sure when, my Maven Eclipse plugin has been upgraded to 2.6 and probably due to some metadata "corruption", its dependencies were not downloaded, running it caused ClassNotFoundException:

[INFO] ------------------------------------------------------------------------
[INFO] Building Whatever
[INFO] task-segment: [eclipse:eclipse]
[INFO] ------------------------------------------------------------------------
[INFO] Preparing eclipse:eclipse
...
-----------------------------------------------------
this realm = app0.child-container[org.apache.maven.plugins:maven-eclipse-plugin:2.6]
urls[0] = file:/home/yclian/.m2/repository/org/apache/maven/plugins/maven-eclipse-plugin/2.6/maven-eclipse-plugin-2.6.jar
urls[1] = file:/home/yclian/.m2/repository/org/codehaus/plexus/plexus-utils/1.1/plexus-utils-1.1.jar
Number of imports: 10
import: org.codehaus.classworlds.Entry@a6c57a42
import: org.codehaus.classworlds.Entry@12f43f3b
import: org.codehaus.classworlds.Entry@20025374
import: org.codehaus.classworlds.Entry@f8e44ca4
import: org.codehaus.classworlds.Entry@92758522
import: org.codehaus.classworlds.Entry@ebf2705b
import: org.codehaus.classworlds.Entry@bb25e54
import: org.codehaus.classworlds.Entry@bece5185
import: org.codehaus.classworlds.Entry@3fee8e37
import: org.codehaus.classworlds.Entry@3fee19d8


this realm = plexus.core
urls[0] = file:/opt/maven-2/lib/maven-2.1.0-uber.jar
Number of imports: 10
import: org.codehaus.classworlds.Entry@a6c57a42
import: org.codehaus.classworlds.Entry@12f43f3b
import: org.codehaus.classworlds.Entry@20025374
import: org.codehaus.classworlds.Entry@f8e44ca4
import: org.codehaus.classworlds.Entry@92758522
import: org.codehaus.classworlds.Entry@ebf2705b
import: org.codehaus.classworlds.Entry@bb25e54
import: org.codehaus.classworlds.Entry@bece5185
import: org.codehaus.classworlds.Entry@3fee8e37
import: org.codehaus.classworlds.Entry@3fee19d8
-----------------------------------------------------
[INFO] ------------------------------------------------------------------------
[ERROR] BUILD ERROR
[INFO] ------------------------------------------------------------------------
[INFO] Internal error in the plugin manager executing goal 'org.apache.maven.plugins:maven-eclipse-plugin:2.6:eclipse': Unable to load the mojo 'org.apache.maven.plugins:maven-eclipse-plugin:2.6:eclipse' in the plugin 'org.apache.maven.plugins:maven-eclipse-plugin'. A required class is missing: org/codehaus/plexus/resource/loader/ResourceNotFoundException
org.codehaus.plexus.resource.loader.ResourceNotFoundException

Tried running Maven with the -cpu argument but it didn't help. maven-metadata-central.xml showed the latest as 2.5.1 and others got 2.6. Fixed by deleting metadata files then re-run Maven.

Bah.

Tuesday, March 31, 2009

RTFL - Read the Fucking Logs

Read the fucking logs? Yes, it happened so many times, on so many guys. Something that I always try to avoid.

Say, you hit into a problem, you looked into the log file, you could see a bunch of error messages and stacktrace, you read the first line, you didn't bother to read the second and further.

You tried to change the configurations or something that you thought would make sense but the same (or new) error came out. You spent quite some time on it like an hour or a day.

You asked me / your supervisor to look into it and they found clear error messages or indications somewhere down the error log. You felt sorry, you found yourself stupid to waste so much time, you wished to be in the hall of shame for that day.

If this is a recurring habit of your, can you make "reading logs carefully" as a troubleshooting principle?

Friday, March 20, 2009

Some Particulars like Room Names in Aflexi

This is truly a random post. During the lunch time at Full House (NZX), I talked about red-tape/bureaucracy in corporates that I can't stand, then exchanges of some Oz slangs.

When I got back to the office, I came across a blog titled "101 Guide to going to the Men's toilet in Sydney" in my ex-company's internal blog (we still maintain a good relationship basically).

Here're a number of things I raised and discussed over in a meeting an hour ago with the team:
  1. The developer's roster, tasks and deadlines.
  2. The right ways to use the toilet.
  3. The ashtray pot and your ciggy.
  4. The cups and basin.
  5. The room names.
Items 2 to 4 are small particulars but people neglect or find themselves resistant in conforming to social standards, if that're the right words to use, anyway. The Aflexi devhub in PJ, Taman Megah Emas (which I named it Aflexi Gold a while ago), has quite a number of rooms and 3 toilets. I'm thinking of giving them names (since 2 months ago but things got me stuck), like country's, city's, etc.

How're rooms named in your office?

Thursday, March 5, 2009

404 on Nexus

Was stuck for a few minutes as Sonatype Nexus was reporting 404 while I tried to access a POM file. Expiring the cache (right clicking on the troubled repository) is the fix of it.

Wednesday, March 4, 2009

Working on Branches and the Stable Trunk

I used to work on the trunk because for all the projects I worked on, it's either I was the sole developer or the team ensured that everyone code check-in has passed the local tests. Otherwise, we created (we still) branches for bug fixes, large code for new features, etc.

About 2 months ago, I had a short git session with Kamal and he shared with me the idea of story/branch. And as my team started to grow with more dev guys, I feel the need of encouraging more branching and now this has become a rule. For a project with automated tests defined with CI set up, a stable trunk is guaranteed and this practice can fit very well into it.

We don't use git but Subversion because that's how we started and we are still comfortable with it now (most of the time).

If you have a working copy on trunk, here're some simple steps to switch to a branch (svn switch won't work):

svn cp https://repo/trunk https://repo/branches/module-version-storyname
cd /path/to/trunk/working/copy
svn diff > /tmp/diff
svn co https://repo/branches/module-version-storyname
cd module-version-storyname
patch -p0 < /tmp/diff

Thursday, February 19, 2009

Wireless Network not working on Ubuntu after Suspend/Hibernate

It never happened to me til about a month plus ago. Didn't bother to fix it 'til just now.

It's a bug reported at 53310, and it's fixed (on my machine) by doing:
modprobe -r ath_pci
modprobe ath_pci

Wednesday, February 18, 2009

Brush up your Support-Fu, can?

Wasted about an hour at the Public Bank HQ this morning, more than 30 minutes on waiting, 2 minutes on getting confused and 5 minutes on getting an answer.

Long story short, yes.

I want to start a long term investment on a mutual fund that I selected and I do not have a saving account in this bank, which you will need to have $ transferred monthly to the investment account. I have my forms, my 500K bank draft (nuh, don't trust this) and my temporary IC (as I lost my purse a few days ago).

There.. reached my turn to be served by this first customer service dude.
Dude A: So you want to open a saving account for that?
Dude A: You will need a minimum of MYR 250. Can I have your IC please?
Me: I do not have an IC.. now. But I can show you my temporary IC.
Dude A: I am sorry sir, we will need that otherwise the system won't pass you.
Me: So does it mean I have to wait 3 weeks before I come here?
Dude A: Yes. I'm sorry.
Me: Options?
Dude A: We need your IC, the system has to scan it.
Me: And there's nothing I can do here anymore today, right?
Dude A: Yes, I'm sorry.
I walked away then paused, and I was annoyed by not giving much details about how exactly it works. I spent my morning to be stuck in the jam, some time to find a parking, waited on the bench, and I got knocked out with a confused mind.

I waited to be served again and this time (by a lady) I got my answer. To start an investment, you do not need a saving account, it's only for the convenience of not needing you to visit the bank to deposit $ for the investment every month. Basically, I could just start it by dumping all my 500K into it, which of course is not my plan. So there're a few options:
  1. Create the investment account, pass them my 500K.
    1. Come back 3 weeks later to open a saving account, OR,
    2. Come back every month to deposit your $.

  2. Wait for 3 weeks to open both accounts and start the investment with your desired amount of $.

I have some Customer Service experience (I worked in both the Atlassian and MuleSource team) and there're a number of things that you have to avoid while handing a ticket / customer.
  1. Never assume that the customers are smart. Like me, I was clueless. So, ask them questions.
  2. Don't jump into solution too fast. Get the context right and make sure that both of you are on the same understanding.
  3. Always give them a solution, if not, workarounds, still not, explain why.
Public Bank, you could lose me as a customer, I still have 3 weeks to think.

Sunday, February 15, 2009

InflaterIjputStream.class

Exception in thread "main" java.lang.NoClassDefFoundError:
java/util/zip/InflaterInputStream
at java.util.zip.ZipFile.getInputStream(ZipFile.java:212)
at java.util.zip.ZipFile.getInputStream(ZipFile.java:180)
at java.util.jar.JarFile.hasClassPathAttribute(JarFile.java:463)
...

In the rt.jar, InflaterInputStream.class was named InflaterIjputStream.class. How could this happen? Fixed the issue by repackaging the zip.

yc

Tuesday, February 10, 2009

JVM failed with Segmentation Fault

This happened completely out of a sudden and truly random. The JVM that runs our business logic app died last night and could never be brought up anymore due to segmentation fault. (It suicided quietly)

The last few lines via strace:

DEBUG DefaultListableBeanFactory [main]: Returning cached instance of singleton bean 'org.springframework.transaction.config.internalTransactionAdvisor'
[02-10 10:22:30] DEBUG AnnotationTransactionAttributeSource [main]: Adding transactional method [createUser] with attribute [PROPAGATION_REQUIRED,ISOLATION_DEFAULT]
) = ? ERESTARTSYS (To be restarted)
+++ killed by SIGSEGV +++

It could just be a JVM bug (not exactly sure which, but suggested by John Raymond Wold, W_work in #java@irc.freenode.net) and upgrading to the newest version fixed the problem.

- yc