|
From: Markus K. <ma...@pr...> - 2016-09-17 21:40:05
|
On 17 September 2016 16:00:02 CEST, "Blum, Jon" <jon...@or...> wrote: >Final update -- I think I've sorted out the problem! > >It was a Wildfly configuration issue -- Wildfly was set up to run as a >service with its own dedicated user, which was not the same user which >was >running the SignServer CLI. > >Basically, I turned on trace logging for the client functions in >conf/log4j.properties (changing the appropriate entries already in the >file >to TRACE level). This revealed that the system was trying to read an >authentication challenge file stored in /opt/wildfly/tmp/auth -- but it >didn't have permission for the directory. As a temporary workaround >I've >changed the folder permissions; in the longer term I'm going to have to >redefine the Wildfly user to put them in the same group. > >So, that point is resolved, and I can get back to proper configuration! > >--jon > > >On Sat, Sep 17, 2016 at 8:57 PM, Blum, Jon <jon...@or...> >wrote: > >> I hope the mailing list has finally accepted my subscription... >> >> Anyway, update: I'm now thinking the problem is most likely with my >> wildfly server configuration. I'd already tried to add an explicit >> remoting connection, and have now removed it, returning to the >default >> configuration... still no luck. At this point the only listener is >the >> default one, on 8080. >> >> So, a few more questions: >> >> 1) There's no way of telling from SignServer's client-side >command-line >> app whether the EJB connection is failing to reach the server, or >whether >> the server is rejecting the request -- right? Is there anything I >should >> be looking for in the server logs when I try to make a connection? >> >> 2) User IDs -- does SignServer require a specific Wildfly user ID to >be >> added to the jboss-ejb-client.properties file? (I've tried it with >and >> without, with no luck -- but it's possible that my user doesn't have >the >> right access. If I need a user, how should I set it up?) >> >> 3) Could the problem be with the following Wildfly commands, which >were >> listed under the JBoss 7 section of the SignServer install doc? I'm >not >> sure whether they apply to Wildfly. Could the presence of >> "jbossws.undefined.host" rather than 127.0.0.1 be causing it to miss >the >> connection? >> >> /subsystem=webservices:write-attribute(name=wsdl-host, >> value=jbossws.undefined.host) >> >> /subsystem=webservices:write-attribute(name=modify-wsdl-address, >> value=true) >> >> Thanks, >> Jon >> >> On Fri, Sep 16, 2016 at 11:07 PM, Markus Kilås <ma...@pr...> >wrote: >> >>> On 09/16/2016 02:36 PM, Blum, Jon wrote: >>> > Hello Markus -- thanks for your guidance here! >>> > >>> >>> (Jon, please subscribe to the list otherwise your responses will not >be >>> delivered to the other subscribers. >>> https://lists.sourceforge.net/lists/listinfo/signserver-develop) >>> >>> >> Just as you explore below the "No EJB recevier available for ..." >>> >> typically means that either SignServer/JBoss hasn't started up >>> >> completely or that the wrong connection configuration is used >from the >>> > CLI. >>> > >>> >> What you want to look for is that you get the EVENT: >SIGNSERVER_STARTUP >>> >> and not any error following it. >>> > >>> > Okay -- I can confirm that Wildlfly and SignServer are starting up >>> > successfully. From the end of server.log: >>> > >>> > 2016-09-16 01:18:48,894 INFO >>> > [org.signserver.server.log.SignServerLog4jDevice] (ServerService >Thread >>> > Pool -- 67) EVENT: SIGNSERVER_STARTUP; MODULE: SERVICE; >ADMINISTRATOR: >>> > StartServicesServlet.init; ISSUER: null; SERIAL_NUMBER: null; >WORKER_ID: >>> > null; msg: start services startup msg; VERSION: SignServer CE >3.7.0; >>> > REPLY_TIME:1474013928894 >>> > >>> > So if the server's running, and the pages are accessible on >>> > localhost:8080, it's got to be something wrong on the client side, >>> right? >>> >>> Yes likely, but it could still be a configuration issue on the >server >>> side for the EJB connector. But if you haven't touch those then I >>> suppose its the client side that is the problem. >>> >>> >>> > >>> > A couple of questions to help me hunt this down -- >>> > >>> > * Are there any other properties files which could affect the >client >>> > EJB connections, besides conf/jboss7/jboss-ejb-client.properties? >If >>> >>> I don't have a WildFly 10 installation at the moment but tested with >>> WildFly 9.0.2.Final and all changes I had to do on the client side >was >>> to edit conf/jboss7/jboss-ejb-client.properties to use port 8080 and >it >>> worked for me. >>> >>> I should also mention that since 3.7.x+ we have changed to instead >>> recommending to turn of the EJB interface on port 8080 and instead >add >>> it back on port 4447 but that is more from a security point as you >can >>> then stop access to the CLI on a network/firewall level just as >before. >>> This change will be available in the 4.0.0 manual which you can have >a >>> sneak peek into here: >>> https://www.signserver.org/sandbox/doc/4.0.0/manual/installg >>> uide.html#WildFly_9_SSL_configuration >>> >>> See the "Securing the CLI by removing...". >>> >>> > I've mis-set something in the build process, could it be including >an >>> > incorrect .properties file at build time? >>> Don't think so. Very few properties has any effect on the build >process. >>> Server-side properties are packaged in the EAR file when you run the >>> bin/ant deploy command and client-side properties like >jndi.properties >>> and the jboss-ejb-thing is read at runtime when running the CLI. >>> >>> > * Is there any way I can check specifically whether *EJB* >connections >>> > are accessible over port 8080? I can confirm the JSP pages are >showing >>> > properly over localhost:8080; but could the beans still be being >>> > blocked? (I did confirm that the beans are mapped to >>> > java:jboss/exported above, so they should be visible... right?) >>> >>> Besides running the CLI (as you did) and the AdminGUI (would give >same >>> error) - not any that I know of. Maybe WildFly has some tool for it? >>> >>> >>> Cheers, >>> Markus >>> >>> >> Great! That was interesting, thanks for sharing. Cheers, Markus -- Sent from my Android phone with K-9 Mail. Please excuse my brevity. |