|Dixie Pine||Oct 7, 2009 3:31 pm||.pdf, .pdf, .pdf|
|Bill Shannon||Oct 14, 2009 8:54 pm|
|Lloyd Chambers||Oct 14, 2009 10:27 pm|
|Byron Nevins||Oct 14, 2009 11:55 pm|
|Dixie Pine||Oct 19, 2009 3:05 pm||.pdf, .pdf, .pdf|
|Bill Shannon||Oct 19, 2009 3:14 pm|
|Lloyd Chambers||Oct 19, 2009 3:25 pm|
|Sankar Neelakandan||Oct 19, 2009 3:38 pm|
|Bill Shannon||Oct 19, 2009 3:45 pm|
|Bill Shannon||Oct 19, 2009 3:51 pm|
|Sankar Neelakandan||Oct 19, 2009 4:09 pm|
|Bill Shannon||Oct 19, 2009 4:41 pm|
|Sankar Neelakandan||Oct 19, 2009 4:56 pm|
|Bill Shannon||Oct 19, 2009 6:11 pm|
|Byron Nevins||Oct 19, 2009 6:46 pm|
|Byron Nevins||Oct 19, 2009 6:57 pm|
|Bill Shannon||Oct 19, 2009 9:02 pm|
|Dixie Pine||Oct 20, 2009 11:11 am|
|Subject:||Re: Review of start-domain, stop-domain, and restart-domainsman pages|
|From:||Byron Nevins (Byro...@Sun.COM)|
|Date:||Oct 19, 2009 6:46:50 pm|
The trivial test results are the same for me. Grizzly is clearly different then simple ServerSocket usage Here is how I reproduce it:
1) have a stock domain1 -- not running 2) create domain2 with a different admin port (e.g. 5050) but the same instance port (8080)  2B) I edit index.html for both domains. Simple test -- add <h1>domain1</h1> or <h1>domain2</h1> respectively 3) start domain1 4) start domain2
On Windows step 4 works with no errors. 
* Now go to a browser and load localhost:8080. * One of the domains will be chosen - usually the second one started -- and it will not switch. You can reload over and over. * Now kill the domain that "owns" 8080. * The browser silently and without any hiccup will switch to using the other domain
I edit index.html for both domains. Simple test -- add <h1>domain1</h1> <h1>domain2</h1> respectively
 asadmin create-domain --adminport 5050 --instanceport 8080 domain2  btrace jvm bind error. Not fatal. I set monitoring-enabled to false to avoid the problem
Bill Shannon wrote:
Lloyd Chambers wrote on 10/19/09 15:26:
See these flags. I do not know what they mean for sure, but they imply re-opening is possible.
I don't know what that's about, but a simple test shows that only a single process can open a server socket on a given port on Mac OS X.
I tried the same test on Windows XP with the same behavior.
I used a trivial test program that does "new ServerSocket(10010)" and then sleeps, then I try to run two of those programs at once.
The second instance fails, as expected, on Mac OS and Windows.
Byron, how did you reproduce this issue?