-
15. Re: Want to help develop a Distributed HA-JMS
bill.burke Mar 14, 2004 2:59 PM (in response to mkprim)If you make absolutely positively sure that you do not break any new tests, and get this in by monday, then put it in head.
-
16. Re: Want to help develop a Distributed HA-JMS
ivelin.ivanov Mar 14, 2004 3:10 PM (in response to mkprim)Bill, I'd rather wait for Marcelo to test the quality before we check in 4-DR3.
Ivelin -
17. Re: Want to help develop a Distributed HA-JMS
ivelin.ivanov Mar 14, 2004 7:15 PM (in response to mkprim)
Per our discussion with Adrian, in 3.2.4 under config "server/all" there is going to be another deployment directory, parallel to "deploy" and "farm", which will host services that are deployed on exactly one node in the cluster.
The tentative name is "deploy-hasingleton". A service under "deploy" declared in deploy-hasingleton-service.xml will deploy the directory in question.
JMS is the first service that will take advantage of this new directory.
Any suggestions how should we name the directory and what other services are good candidates for it?
Cheers,
Ivelin -
18. Re: Want to help develop a Distributed HA-JMS
ivelin.ivanov Mar 16, 2004 12:40 AM (in response to mkprim)I am starting to check in. The following SAR can be used for manual tests and example HA JMS client side code.
http://cocoonhive.org/jboss/hajms-3.2.4-examples.sar.zip
Ivelin -
19. Re: Want to help develop a Distributed HA-JMS
ivelin.ivanov Mar 16, 2004 2:04 AM (in response to mkprim)Code checked in CVS, branch 3.2. Please help me test it further. I've tested one server with HSQL and two servers with shared MySQL db.
I could use some help with updating the HOW-TO document and moving it to JBoss Wiki.
http://cocoonhive.org/articles/jboss/20031125/JBoss32-hajms.html
Ivelin -
20. Re: Want to help develop a Distributed HA-JMS
mkprim Mar 18, 2004 1:37 PM (in response to mkprim)Sorry I was off these days... I'll test it today...Tomorrow I'll post the answer.
Maybe you would like me to post the test code? Or to mail it to you or upload it to cvs? -
21. Re: Want to help develop a Distributed HA-JMS
ivelin.ivanov Mar 20, 2004 10:18 PM (in response to mkprim)Let's use this forum for now. When the code is tested and we know it's working, then we'll post the examples to the Wiki:
http://jboss.org/wiki/Wiki.jsp?page=JBossMQHA
Ivelin -
22. Re: Want to help develop a Distributed HA-JMS
ivelin.ivanov Mar 21, 2004 5:43 AM (in response to mkprim)I populated the document on Wiki. Feel free to add more usage scenarios and examples.
http://jboss.org/wiki/Wiki.jsp?page=JBossMQHA
Ivelin -
23. Re: Want to help develop a Distributed HA-JMS
mkprim Mar 23, 2004 2:25 PM (in response to mkprim)How can I add my example code to the JBoss Wiki?
Do you want me to mail it to you Ivelin?
Congratulations and Thanks,
Marcelo -
24. Re: Want to help develop a Distributed HA-JMS
ivelin.ivanov Mar 24, 2004 1:07 AM (in response to mkprim)You should be able to create a new account for wiki and then add your code either inline or attach it.
Ivelin -
25. Re: Want to help develop a Distributed HA-JMS
acoliver Aug 11, 2004 10:24 AM (in response to mkprim)It seems to be a flaw to me that we have to choose either to have HAJMS or non-HAJMS. I ought to be able to have both. Meaning if I look up via HAJNDI, I get the "master node" JMS. IF I look up via JNDI, I get my local JMS. Furthermore, It seems like I ought to be able to set up local topics/queus to be "clients" to the "master" topics/queues and that the local topic/queue ought to be able to have "selectors" for subsets.
-
26. Re: Want to help develop a Distributed HA-JMS
adrian.brock Aug 11, 2004 11:24 AM (in response to mkprim)local or HA is a configuration issue.
There's nothing to stop you deploying JBossMQ twice.
I'd like to see the HA choice more fine grained at the
queue/topic subscription level, but I can only repeat what
I've said many times before...
Explain how you elect one server to be the transaction
controller for each session and how you coordinate the disparate work
efficiently without turning your whole cluster into a single threaded
machine.
Or similarly, how to make a contested queue work across a cluster
with an efficient distributed lock (if such a thing even exists)
or equivalent mechanism.
Selectors on queues are usually bad.