users@glassfish.java.net

Re: output truncate bug

From: <forums_at_java.net>
Date: Sat, 17 Mar 2012 09:16:30 -0500 (CDT)

For me, stability is always more important that raw performance. In real
world performance depends on backend. Application server (unless you are
using EJB, especially message driven) is unlikely to be performance problem.

I had numerous stability problems with glassfish, output truncate bug is just
one:

  1) hangs on long lived AJP connections without traffic
  2) JNDI resource resolving is wrong (or at least unexpected) for custom JNDI
     resources -  [1] GLASSFISH-17640 [2]
  3) Domain -> instance synchronization problems
  4) It suddently stops working and start hanging even without configuration
     changes and restart didnt helped.
  5) connection leak detection is not instant but delayed - GLASSFISH-17528
     [3]
  6) after stressing gf hangs, it cures itself after few minutes but since
     then runs very slowly
  7) Flush pool does not work - [4] GLASSFISH-17573 [5]
  8) Eclipse plugin for deployement during development didnt worked properly

I collected some jstack dumps from hanged gf and i will post them here.

But my main point is that i do not want to spend my time on debuging and
improving GF because i have other work to do. I develop applications and not
want to spend my time on backend stuff unless is really neceseary. I spended
some time on Apache Cassandra, because there was no suitable alternative
(HBase sucked). i expect backend servers (like JMS broker, SQL database and
application server) to do their stuff correctly.

 


[1] http://java.net/jira/browse/GLASSFISH-17640
[2] http://java.net/jira/browse/GLASSFISH-17640
[3] http://java.net/jira/browse/GLASSFISH-17528
[4] http://java.net/jira/browse/GLASSFISH-17573
[5] http://java.net/jira/browse/GLASSFISH-17573

--
[Message sent by forum member 'rkolar02']
View Post: http://forums.java.net/node/884043