Push not working inside GF Virtual Server web app

[mauritzl]

Scenario:

We have one glassfish instance that is configured to answer to multiple 'virtual hosts', e.g appserver.ourcompany.com, billing.ourcompany.com, crm.ourcompany.com. We do not run Apache in front, since we are not facing any scalability issues and we only run JSF applications that scale well when accessed directly. Each of the virtual servers / hosts answer to different JSF web applications. Many of the web applications use PrimeFaces (and hopefully Push).
We have developed push application locally and everything seems to work fine, until we deploy the web applications to their virtual server targets. Then push stops working, typically emitting errors like this:

[2014-02-12T18:32:51.907+0100] [glassfish 4.0] [WARNING] [] [org.atmosphere.cpr.MetaBroadcaster] No Broadcaster matches /progressmessages. Message {"data":"progress"} WILL BE LOST. Make sure you cache it or make sure the Broadcaster exists before.

Seems to us like this is same issue as many have experienced when using Apache httpd VirtualHost in front of GF / Tomcat / other JEE container.

Have tried many combinations of primefaces.PUSH_SERVER_URL, like 'http://crm.ourcompany.com:8282', but it doesn't seem to help the WebSocket client / server communication to work.

Any ideas?

[smithh032772]

Problem running an application in Glassfish cluster

In that discussion, Aleksandar Nikolov is asking the question, and he (usually) 'always' provides a fix/workaround to make Atmosphere work with his target environment.

I recommend that you follow his contributions in the forum, and try any/everything that he is doing and/or recommending, and 'after' you read any/every discussion that he is participating in, and you are not able to solve your issue, then i recommend that you post a mail on the atmosphere google groups mail list. you may get a response from Jeanfrancois or Aleksandar Nikolov.

[mauritzl]

Thank you for responding.

The issue regarding running the web application inside a Glassfish Cluster is somewhat different since it seems to deal with a problematic instance ID appended to the session cookie causing trouble, but I did not get the feeling that Nikolov was running the web applications in GF 'Virtual Server' configurations. Nice to know about the clustering issue anyway, since we might bump into that one later.

I will search further and might post to the atmosphere group if I am unable to find a solution.

A temporary workaround is to resort to the polling component (p:poll) to have some kind of progress bar possibility (which is currently what we use push for) ... .

[smithh032772]

have you searched this forum for PUSH_SERVER_URL ?

i'm about to search for you. one guy, here, posted a question and a workaround related to PUSH_SERVER_URL.

[smithh032772]

have you searched this forum for PUSH_SERVER_URL ?

i'm about to search for you. one guy, here, posted a question and a workaround related to PUSH_SERVER_URL.

primefaces.PUSH_SERVER_URL with different urls

read that from start to finish. please do not reply and say, it is not glassfish in that topic, so it cannot help. i think it may be to help.