After reading an excellent article on Asynchronous processing support in Servlet 3.0 at JavaWorld.com, I have one more reason to use Servlet 3 providers. Just as Jetty Asynchronous REST example shown, real world requets that aggregates several backend call responses benefit from the asynchronous nature by freeing valuable CPU resources, improving server scalability.
Try this with PHP 
Kudos for the JSR expert group for making this a standard!
In my current job I’m working with some PHP codes for which I’ve implemented the singleton pattern. My testcase is quite straightforward and consists of a singleton that stores the time at which it was instantiated:
class MySingleton {
private static $myVar;
private static $instance;
private function MySingleton() {
self::$myVar = time();
}
public static function getInstance() {
if (! isset(self::$instance)) {
self::$instance = new MySingleton();
}
return self::$instance;
}
public function getMyVar() {
return self::$myVar;
}
}
echo "Time: " . MySingleton::getInstance()->getMyVar();
Now, let’s test it. This is the result for the first invocation:
Time: 1274882765
and now a second invocation:
Time: 1274882775
So what’s going on here? PHP architecture is designed to fork a new process on every page request. That means that the singleton only lasts for a request. On the next request, it is instantiated again, that’s why we get different timestamps. What if we want for the singleton to span several requests, i.e. a session? Here is a modified version:
class MySingleton {
private static $myVar;
private static $instance;
private function MySingleton() {
self::$myVar = time();
}
public static function getInstance() {
if (! isset(self::$instance)) {
self::$instance = new MySingleton();
}
return self::$instance;
}
public function getMyVar() {
return self::$myVar;
}
}
session_start();
if (! isset($_SESSION["mySingleton"])) {
$_SESSION["mySingleton"] = MySingleton::getInstance()->getMyVar();
}
echo "Time: " . $_SESSION["mySingleton"];
The result is weird:
Time:
There’s no output because I used “static” scope for $myVar in MySingleton and PHP won’t serialize it. So I must remove the “static” keyword to make it an instance variable:
myVar = time();
}
public static function getInstance() {
if (! isset(self::$instance)) {
self::$instance = new MySingleton();
}
return self::$instance;
}
public function getMyVar() {
return $this->myVar;
}
}
session_start();
if (! isset($_SESSION["mySingleton"])) {
$_SESSION["mySingleton"] = MySingleton::getInstance();
}
echo "Time: " . $_SESSION["mySingleton"]->getMyVar();
and now it works wonderfully thru all invocations… well, at least for the same session. If there are multiple sessions, each of them will have their own singleton so at any point in time, there will be N “singletons” given N clients.
In conclusion
PHP singletons are request scoped, or at most session scoped. However, it might not worth the effort to span its scope to session since PHP has to serialize sessions to disk, memcached, or whatever mechanism used for session serialization, and it might lead to an I/O problem. You should weight the cost of instantiation of objects and session serialization. In general, I would recommend session variables as an exception. Try to externalize states by embedding them to the URL (such as the current page in a paginated request) and reserve session variables for simple variables such as page hits counter in a session fashion.
While integrating swirstatdsample from sourceforge.net to retrieve remote host system stats I bumped into an interesting problem which is that rstatd sampler will run forever until we shut it down. So, this make it unsuitable for continuous builders like Hudson that run Performance Plugin to obtain JMeter performance trends. To work around this, I needed to stop the test when JMeter active thread count is equal to 1, i.e. only the rstat monitor should be running. After searching for almost a day I found this from JMeter user manual seccion 19.5.7 “__javaScript”:
Rhinoscript allows access to static methods via its Packages object. See the Scripting Java documentation. For example one can access the JMeterContextService static methods thus:
Packages.org.apache.jmeter.threads.JMeterContextService.getTotalThreads()
Thus in JMeter the total threads active at any given time is available at JMeterContextService. You can either use the __javaScript function as explained above or use BeanShell Sampler which allows a more Jav-like syntax:
activeThreadCount = org.apache.jmeter.threads.JMeterContextService.getTotalThreads();
Today I had the pleasure (or pressure) to install RSTATD on Ubuntu since we needed to monitor one of our hosts during a stress test. It turned out to be not so simple so I’m documenting the steps here:
sudo xinetd rstatd
We need to specifically ask for xinetd since we need to configure RSTATD service below.
# default: off
# # description: An xinetd internal service which rstatd’s characters back to clients.
#
service rstatd
{
type = RPC
rpc_version = 2-4
socket_type = dgram
protocol = udp
wait = yes
user = root
log_on_success += USERID
log_on_failure += USERID
server = /usr/sbin/rpc.rstatd
disable = no
}
sudo apt-get install rstat-client
Today I was trying to access a Samba file server in my network when I run into the issue that a symbolic link is retrieved “as is” from the Samba server instead of following it. It puzzled me at first since I thought that this was a server configuration but it turns out that the client can decide on it as well.
smbmount //server/share /mntpoint -o username=user password=pass nounix
Note the nounix option which does the trick to follow symbolic links.
And if you need this drive to be mounted upon start you can add an entry in /etc/fstab:
//sambaserver/clee /mnt/sambaserver/clee cifs credentials=/root/.smbcredentials,dir_mode=0777,file_mode=0777,nounix 0 0
Where /root/.smbcredentials contains your user and password in this format:
username=your_username password=your_password
Remember to change the permissions also:
sudo chmod 600 /root/.smbcredentials
Sources:
http://www.guia-ubuntu.org/index.php?title=Montar_comparticiones_remotas_usando_smbfs_y_cifs
You started your GoogleCode project but don’t know how to set your project up with Hudson? If so, read on.
First you should be aware that SVN client caches your credentials. So in the machine where Hudson runs, log in as the user that runs Hudson and issue a SVN command against your repository as instructed here:
http://blogs.open.collab.net/svn/2009/07/subversion-16-security-improvements.html
To install the plugin go to to Hudson Dashboard -> Manage Hudson -> Manage Plugins. Under the Available tab, find the GoogleCode plugin and click on install.
Go to Hudson Dashboard -> New Job. Enter the project name and click on “Build a maven2 project”. Click on ok. Enter the Google code website which should be:
http://code.google.com/p/<project name>/
For the source management section, please note that there is a bug in GoogleCode plugin that will fail to authenticate the developer connection with GoogleCode so please follow the workaround here to complete your setup:
http://malsserver.blogspot.com/2009/04/hudson-maven-release-plug-in-plug-in.html
And that’s it! Happy coding!
The reason to do this is that all changes to trunk should be traceable from day zero and by branching we lose the revision history. The purpose of a branch is generally to create a release and apply patches to it, thus it makes sense to isolate the revision history from the trunk since we already know the changelist at which we branched from trunk. The same concept applies to development branches.
Certain kinds of changes like maven-ization require moving files around to comply with the Maven convention though it’s best to override the convention by specifying multiple resource directories and at a later time divide them into multiple modules to conform a Maven multi-module project. But in case you must move them because of other reasons (e.g. you mistyped the name), create a branch spec that maps the original location of the moved elements to the new location so that you can integrate changes from the trunk until you finish your work in your branch. Note that SVN won’t detect which revisions have already been integrated unlike Perforce, so you are URGED to put the integrated revisions in your commit message to keep track of integrated revisions.
I think most of us are aware of the benefit of making an Eclipse project A depend on another project B directly, instead of depending on the project B jar. Just in case you don’t, consider the scenario where project A is a WAR application and B is a JAR containing your domain objects. Then you deployed it using WTP server support or Sysdeo Tomcat plugin to test it locally. The difference is the following:
In order to make your multi-module maven project declare dependencies between the modules of the project, you should use <scope>compile</scope> and not any other combination (such as “provided” as I’ve seen in some production code). This way, Eclipse will recognize the dependency as a Eclipse project and won’t include it as a JAR
Somewhere in the time, I’ve updated to a newer Live Messenger and stopped receiving new mail notificacions (the familiar popup windows saying “You’ve received a new mail from XYZ”). I’ve checked the option to notify me of new mails in Tools -> Settings without luck:
The "Display alerts when e-mail is received" option is checked
What’s more, the new mail count does not update after messages are read as it used to behave. Now, you have to sign out of Live Messenger and log back in again to have them refreshed. I talked to a bunch of friends running Live Messenger and the problem seems to be widespread; there are people running Live Messenger 2008, 2009 and a lot of them where having problems.
The new mail count does not update at all after mails being read
Having that said, I decided to report a bug to Microsoft but Google is not telling me the place where I could report it. So, I’m posting here until someone points me to the right URL.
Well, while writing another post I found out another missing feature for Blogger. It seems that I’m not giving Blogger a break after all
Have you ever tried to add an internal link to your post? Consider a really long post:
This is a loooooong post which links to paragrah 3 in this post.
Paragraph 1…..
Paragraph 2…..
Paragraph 3 !!!
Try to do the same in Blogger
WordPress 7 – Blogger 1
