DDevelopment & Design

Solving package dependency problem using synonyms

Problem I take part in supporting an ODS system. One of the biggest problem during deployment of revisions is the dependency between packages. Suppose a package A has to be changed and that it is referenced be two packages: B1 and B2 which are used in some database sessions. If the the A package is altered and it contained some global variables then the next execution of the packages B1 or B2 in those sessions triggers “ORA-04061: existing state of package has been invalidated” error. This is because global variables preserve their values during database sessions. Alteration of the package code may change the variables, delete some of them, add others. Therefore, they have to be initialized again. Reseting the global variables during the session could lead to inconsistency. In order to prevent this. Oracle throws the ORA-04061 exception. This mechanism is described in details on the page 

http://oraclequirks.blogspot.com/2007/03/ora-04061-existing-state-of-package-has.html

In addition, alteration of the package body or package specification is not possible as long as it used in a session. The replacing package DDL (CREATE OR REPLACE PACKAGE [BODY]) waits until no session uses the package. This may never happen – if new session starts using package before the previous one stops doing that. In order to change the package – all sessions using it must be killed. It is quite common that a session cannot be killed right now, but is moved into MARKED FOR KILLED state. As a result, the deployment of new package version is postponed for indefinite period of time until the session is really killed. I even created a script which kills all sessions which blocks any session started from my computer by my OS user. The script is really crude but it works and is really useful so I am posting it here.

DECLARE
  CURSOR c_sql_hist(pp_machine VARCHAR2, pp_osuser VARCHAR2) IS
    SELECT ash.blocking_session, ash.blocking_session_status, ash.blocking_session_serial#, s.program, s.module, s.action
      FROM v$active_session_history ash
      JOIN v$session s ON ash.blocking_session = s.sid
                      AND ash.blocking_session_serial# = s.serial#
     WHERE ash.session_id in ( select sid from v$session where machine = pp_machine
     and osuser = pp_osuser )
       AND ash.sample_time > SYSDATE - 2 / 24 / 3600
     ORDER BY sample_time DESC;
  v_r c_sql_hist%ROWTYPE;

  v_my_machine v$session.MACHINE%TYPE;
  v_my_osuser v$session.osuser%TYPE;
BEGIN
  select machine, osuser into v_my_machine, v_my_osuser from v$session
  where sid = ( Select Sid from v$mystat where rownum = 1);

  for i in 1..10000 loop  -- just to prevent infinite execution
    dbms_output.put_line('step: ' || i );
    OPEN c_sql_hist(v_my_machine, v_my_osuser);
    FETCH c_sql_hist
      INTO v_r;
    IF NOT c_sql_hist%NOTFOUND THEN
      dbms_output.put_line(v_r.blocking_session || ' ' || v_r.blocking_session_status || ' ' || v_r.blocking_session_serial# || ' ' || v_r.program || ' ' ||
                           v_r.module || ' ' || v_r.action);
      execute immediate ' ALTER SYSTEM KILL SESSION ''' || v_r.blocking_session || ',' || v_r.blocking_session_serial# || '''';
    else
      exit;
    END IF;
    CLOSE c_sql_hist;
    dbms_lock.sleep(0.5); -- sleep is used to get new blocking sessions
  end loop;
END;

In order to reduce the dependency problems described above, various tool packages, which should be located in one central schema, were copied to all schemas. As a result, they may be changed without killing all (only some) processes. However, this solution is terrible from the point of view of code maintenance.

Solution

The solution to this problem is to reference packages by synonyms. If the package need to be changed – a new version of it is installed with a different name. After it is tested – the synonym is altered to point the new version. The nearest execution of this synonym within a session causes the new package to be executed. Its global variables are initialized, their previous values are lost but this is not a problem in our ODS as they work as a cache and does not hold any state between executions. There is an additional limitation that one procedure or package cannot be executed concurrently is its instances use different object through the same synonym. Suppose that after B1 is executed, the A synonym is changed to point to A_v2. If B1 is executed again (directly or indirectly) then the second execution of B1 will not start before the old execution using the old package is finished. However, this is not really a problem. The solution described above allows to deploy database code efficiently. It is however necessary to remember how this mechanism works and what the limitations are. A naming convention needs also to be developed (for example adding the result of TO_CHAR(sysdate, ‘$MM_DD’), which is dollar sing, month and day of deployment, to the package name). It is also necessary to give other schemas the same grant to new package version as were given to the old package version.

— Previous article

<!--:en-->Complex flows with Apache Camel<!--:-->

Next article —

<!--:en-->Accessing Microsoft SQL Server from Oracle<!--:-->

You May Also Like
DDevelopment & Design

Grails session timeout without XML

This article shows clean, non hacky way of configuring featureful event listeners for Grails application servlet context. Feat. HttpSessionListener as a Spring bean example with session timeout depending on whether user account is premium or not.

Common approaches

Speaking of session timeout config in Grails, a default approach is to install templates with a command. This way we got direct access to web.xml file. Also more unnecessary files are created. Despite that unnecessary files are unnecessary, we should also remember some other common knowledge: XML is not for humans.

Another, a bit more hacky, way is to create mysterious scripts/_Events.groovy file. Inside of which, by using not less enigmatic closure: eventWebXmlEnd = { filename -> ... }, we can parse and hack into web.xml with a help of XmlSlurper.
Even though lot of Grails plugins do it similar way, still it’s not really straightforward, is it? Besides, where’s the IDE support? Hello!?

Examples of both above ways can be seen on StackOverflow.

Simpler and cleaner way

By adding just a single line to the already generated init closure we have it done:
class BootStrap {


    def init = { servletContext ->    

        servletContext.addListener(OurListenerClass)    

    }    

}

Allrighty, this is enough to avoid XML. Sweets are served after the main course though :)

Listener as a Spring bean

Let us assume we have a requirement. Set a longer session timeout for premium user account.
Users are authenticated upon session creation through SSO.

To easy meet the requirements just instantiate the CustomTimeoutSessionListener as Spring bean at resources.groovy. We also going to need some source of the user custom session timeout. Let say a ConfigService.
beans = {    

    customTimeoutSessionListener(CustomTimeoutSessionListener) {    

        configService = ref('configService')    

    }    

}

With such approach BootStrap.groovy has to by slightly modified. To keep control on listener instantation, instead of passing listener class type, Spring bean is injected by Grails and the instance passed:
class BootStrap {


    def customTimeoutSessionListener


    def init = { servletContext ->    

        servletContext.addListener(customTimeoutSessionListener)

    }    

}

An example CustomTimeoutSessionListener implementation can look like:
import javax.servlet.http.HttpSessionEvent    

import javax.servlet.http.HttpSessionListener    

import your.app.ConfigService    


class CustomTimeoutSessionListener implements HttpSessionListener {    


    ConfigService configService


    @Override    

    void sessionCreated(HttpSessionEvent httpSessionEvent) {    

        httpSessionEvent.session.maxInactiveInterval = configService.sessionTimeoutSeconds

    }    


    @Override    

    void sessionDestroyed(HttpSessionEvent httpSessionEvent) { /* nothing to implement */ }    

}
Having at hand all power of the Spring IoC this is surely a good place to load some persisted user’s account stuff into the session or to notify any other adequate bean about user presence.

Wait, what about the user context?

Honest answer is: that depends on your case. Yet here’s an example of getSessionTimeoutMinutes() implementation using Spring Security:
import org.springframework.security.core.context.SecurityContextHolder    


class ConfigService {


    static final int 3H = 3 * 60 * 60

    static final int QUARTER = 15 * 60


    int getSessionTimeoutSeconds() {    


        String username = SecurityContextHolder.context?.authentication?.principal    

        def account = Account.findByUsername(username)    


        return account?.premium ? 3H : QUARTER

    }    

}
This example is simplified. Does not contain much of defensive programming. Just an assumption that principal is already set and is a String - unique username. Thanks to Grails convention our ConfigService is transactional so the Account domain class can use GORM dynamic finder.
OK, config fetching implementation details are out of scope here anyway. You can get, load, fetch, obtain from wherever you like to. Domain persistence, principal object, role config, external file and so on...

Any gotchas?

There is one. When running grails test command, servletContext comes as some mocked class instance without addListener method. Thus we going to have a MissingMethodException when running tests :(

Solution is typical:
def init = { servletContext ->

    if (Environment.current != Environment.TEST) {    

        servletContext.addListener(customTimeoutSessionListener)    

    }    

}
An unnecessary obstacle if you ask me. Should I submit a Jira issue about that?

TL;DR

Just implement a HttpSessionListener. Create a Spring bean of the listener. Inject it into BootStrap.groovy and call servletContext.addListener(injectedListener).
DDevelopment & Design

Visualizing GIS data in JavaFX 2.0 beta using GeoTools

Geographic data mostly comprises of polygon coordinates sets along with attributes, like country or city name, etc. This is quite easy to visualize in JavaFX, which supports rendering for SVG paths. In the article, I show how to read such GIS data from...Geographic data mostly comprises of polygon coordinates sets along with attributes, like country or city name, etc. This is quite easy to visualize in JavaFX, which supports rendering for SVG paths. In the article, I show how to read such GIS data from...