Re: Ask again: how Sandesha works

[Aleksander Slominski]

Jian J Zhang wrote: I sent out a note asking how Sandesha works. Someone of the list replied but I deleted the reply email by mistake before I had had a chance to read it. I could not find the reply in the archive either. Can the person who replied reply again? I really apprciate it. -------------------------- The original note Here I'd like to understand how Sandesha works.Your help is appreciated. As far as I understand, Sandesha saves the request first in the persistence and then sends out the acknowledgement. Then it invokes the web service (the application code).  The reliability is guaranteed because the message is saved and can be resent to the web service if needed, just like messaging system. Here is what I do not understand:     Assume that the sandesha successfully saves the message into the database (database transaction is committed) and sends out the acknowledgement,     while before it forwards the message to the web service, the server crashes.     Now we restart the server and AXIS plus Sandesha are ready to serve new requests.  But how about the saved request?    AXIS  is only invoked when we have a HTTP/SOAP message. If  we do not get new request, then the message will never be reforwarded again? hi, I believe that Sandesha currently does not support this scenario: when server is restarted it should read queue of messages undelivered to services and restart delivery pipeline. however this was discussed and hopefully will be possible to implement in AXIS2 as main problem is that in AXIS1.2x it is very hard to control message processing pipelines. also i think currently Sandesha is not using database to store messages (by default?) however details on this must be provided by Jaliya as he is implementing it. unfortunately the wiki pages for Sandesha disappeared during chinese wiki spam wars ... http://nagoya.apache.org/wiki/apachewiki.cgi?SandeshaProjectPages/Changes so you need to rely on mail archives http://news.gmane.org/gmane.comp.apache.webservices.fx.devel if you want to find more about Sandesha design and implementation thanks, alek -------- Original Message -------- Subject: RE: Sandesha --(advise needed) Date: Wed, 13 Oct 2004 10:01:47 +0600 From: Jaliya Ekanayake <[EMAIL PROTECTED]> To: 'Krishnendu Kunti' <[EMAIL PROTECTED]> CC: <fx-dev@ws.apache.org> Few more ideas for you..   In current Sandesha architecture the idea was to make both queue and database options available (queue may be used when there is no database available). So writing a database code in the ServerQueue class will not be good. What we should do is implementing the class IServerDAO (or fill ServerDatabaseDAO).   So if you can have an implementation for each function of IServerDAO then the storing happens in a database. To keep the database connection we can use some static variables in ServerDatabaseDAO. And also an init() method can be added to IServerDAO for initiation purposes.   Thanks,   Jaliya and Chamikara     From: Krishnendu Kunti [mailto:[EMAIL PROTECTED]] Sent: Tuesday, October 12, 2004 7:35 PM To: Jaliya Ekanayake Cc: fx-dev@ws.apache.org Subject: Sandesha --(advise needed)   Hi Jaliya   I have created class that opens connection to database (reading some user defined URL, PWD, USERNAME and DRIVER MANEGER—at present I gave these values in the Constants class), maintains connection pool and stores a set of prepared statements which will be used by all the clients. This class needs to be in memory (runnig ) as the connections, prepared statements will be initiated only at the startup.   Is the ServerQueue class package, “org.apache.sandesha.server.queue” persisted across two distinct client calls , if yes then can I assume that my class will be also persisted once it is deployed. If no then how to persist/ (something like deamon) for the purpose of maintaining the connection pools, prepared statements.   Is there any class which takes care of running deamons, so that I might start this class along with it.   Regads Krish   A man with a dream will not be denied   Krishnendu Kunti   SETLABS, Infosys Technologies LTD Bangalore, India Fax: +91 80 51102520 Mobile: 9886375255 URL: http://setlabs/wscoe (Infosys Intranet)       -- The best way to predict the future is to invent it - Alan Kay -------- Original Message -------- Subject: RE: configuring mysql database for Sandesha Date: Thu, 7 Oct 2004 14:30:04 +0600 From: Jaliya Ekanayake <[EMAIL PROTECTED]> To: 'Krishnendu Kunti' <[EMAIL PROTECTED]>, <fx-dev@ws.apache.org> Hi,   As I have replied you, Currently Sandesha is not using any database. It is using a in-memory queue. RMSequence class is handling this queue functionality. We are in the process of modifying Sandesha so that It will handle the messages better. And also with the new architecture Sandesha will improve its interoperability with other WS-Extentions.   We will be putting documentation for the new Sandesha soon. At the moment you can see the architecture using the following wiki. http://nagoya.apache.org/wiki/apachewiki.cgi?SandeshaProjectPages/Changes   Thanks,   Jaliya     From: Krishnendu Kunti [mailto:[EMAIL PROTECTED]] Sent: Wednesday, October 06, 2004 1:45 PM To: fx-dev@ws.apache.org Subject: configuring mysql database for Sandesha     Hi   IS there a configuration file in Sandesha for pointing to mysql database? Which config file / class file should the location of the database be mentioned. Guys can anyone share a more detailed documentation or any link on the net which does the same. Regards Krish   A man with a dream will not be denied   Krishnendu Kunti   SETLABS, Infosys Technologies LTD Bangalore, India Fax: +91 80 51102520 Mobile: 9886375255 URL: http://setlabs/wscoe (Infosys Intranet)     -------- Original Message -------- Subject: Re: Getting Started with Sandesha Date: Sat, 25 Sep 2004 06:36:40 -0400 (EDT) From: [EMAIL PROTECTED] To: Dave Everson <[EMAIL PROTECTED]> CC: fx-dev@ws.apache.org References: <[EMAIL PROTECTED]> Hi Dave, I will try to explain some suggestions for the scenario you mentioned as follows. According to the usage of the WS-RM it will be beneficial for us to first identify the reliability of the message before going in to any processing. That is if we can identify whether the incoming request is a new message or a re-transmission of a previous message we can easily ignore the latter. However there is a problem in this approach. That is when we try to use WS-RM with WS-Security(encryption, digital signatures) then some one can encrypt the WS-RM headers as well. So the message should first pass through the Security handler and then only it can pass through the WS-RM. But as you have explained if we can use our own encryption/decryption mechanism for the payload then we can easily do it after the reliability checks as long as it will not encrypt the SOAP headers. A further enhancement will be that, we can just decrypt the reliability headers and then check the WS-Reliability first and decrypt the rest of the message only if that is passed. However this scenario will not be achieved directly using the current axis implementation. That is coming back to a handler that has already passed. But by deploying custom handlers you may still able to achieve this. WS-RM mainly deals with SOAP headers and it may not require the body of the message (payload) at all. So it will be beneficial for us to do the verification of the payload against a schema the reliability checks. Otherwise the engine will waste its time validating a duplicate message or a re-transmission. At the moment the WS-RM support we have in axis (Apache Sandesha) works at the provider (pivot) level and works at the last stage. This is because we need to provide INORDER EXACTLY ONCE delivery of messages to the service. That is the service is invoked using the original message sequence and not by the order in which they are received. If the quality of service is something other than INORDER that is just EXACTLY ONCE or by the order they are received, then we may be able to modify Sandesha and use it before the other handlers that deal with the payload decryption and validation. We are in the process of modifying Sandesha so that it will improve its usage. We may be able to provide that functionality in the near future and it will not affect the usage of axis other than the way in which the services are deployed. Hope I have explained my suggestions clearly. If you need further clarifications, just send a mail.. Thanks, Jaliya > Hello - > > We are starting a new web service project using Axis as our web service > engine. The requirements for our web services include reliable messaging, > digitial signature, encryption of payload, and validation of XML in > payload > against schemas housed in the XML registry on our server. > > We are considering using several components of WS-FX to accomplish several > of these requirements. > > We were thinking that many of the above requirements could be handlers > with > Axis Handlers before the message even gets to our true web service. > > If the inbound message failed the checks (e.g. digitial signature, > validation against schema), we would throw a SOAPFault and stop processing > the message. > > When I throw reliable messaging into the mix, where is the proper place > for > this? Should I do this as the first step in the process (i.e as soon as > we > get a message) or after all the checks have been done with the message? > > What's the suggested approach based on your experience? What's the pros > and > cons of doing WS-RM functionality before or after the other processing? > > Any recommendations that you can provide would be greatly appreciated. > > Thanks! > > _________________________________________________________________ > FREE pop-up blocking with the new MSN Toolbar – get it now! > http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/ > > >