KnownissuesMessaging

From MRG

Jump to: navigation, search

1 Red Hat Enterprise MRG - Known MESSAGING Issues
2 MRG Management Console Preview
3 Java Clients
4 General
5 Notes
6 Issues in other MRG Components

[edit]

Red Hat Enterprise MRG - Known MESSAGING Issues

[edit]

MRG Management Console Preview

Thank you for evaluating our preview release of the MRG management web UI.

There are the following know issues with the MRG management preview:

- Group remove navigates to a blank page instead of returning to group list page
- The user is not prevented from closing or detaching the session that connects cumin to the managed broker
- Broker IDs are currently not stable across broker restarts

The following instructions assume you are running as root, unless otherwise indicated:

[edit]

1. Install packages

First, use yum to install the management packages. If you haven't already, you may want to install the messaging packages as well.

    # yum install mrg-management
    # yum install mrg-messaging

[edit]

2. Prepare the database

In order to access the database, you must change the authentication method of the postgresql database:

    # vi /var/lib/pgsql/data/pg_hba.conf

For each local connection line, change "ident sameuser" to "trust".

Now you can start the database:

    # service postgresql start

Load the cumin database and schema:

    # cumin-database-init

[edit]

3. Start a managed broker

To make your local broker manageable, edit /etc/rhmd.conf and add

  "mgmt=yes" on its own line.  Then restart the broker so this change takes effect.

[edit]

4. Start cumin (cumin is the name of the web console)

As root:

    # cumin

As an unprivileged user:

    $ cumin --port 8080 # On an unprivileged port

In a browser, go to localhost (or localhost:8080). You should be able to click "Register Brokers" and add brokers by their address and port.

[edit]

Java Clients

It should be noted that Java 1.5 or Java 1.6 is required to use the Java clients.

The java client library uses a logging facade called SLF4J from http://www.slf4j.org/ We do not impose a logging system on your project. SLF4J allows you to redirect the output to your preferred logging system.

Examples include.

No loggin - slf4j-nop.jar
Console - slf4j-simple.jar
Log4J - slf4j-log4j12.jar
Java 1.4 loggin - and slf4j-jdk14.jar
Provide your own adapter.

In order to run you program you need to have one of these jar files in your classpath.The rhm-docs rpm has the slf4j-simple.jar (console output) in order for the examples to work. You can download the others from ttp://www.slf4j.org/ or you can provide your own custom adapter for you application.

The java client connection URL format is described on: http://cwiki.apache.org/confluence/display/qpid/0.10+Connection+URL+Format

It should be noted that (when JNDI is not used) the ConnectionFactory class to be used is: org.apache.qpid.client.AMQConnectionFactory

It should be noted that connection interruption events are not yet propagated to the connection exception listener.

[edit]

General

This page provides updated information about known issues in the MESSAGING component of the Red Hat Enterprise MRG beta. The information is provided in order to raise awareness of known problems so that beta participants can be most efficient and not waste time. This information will be updated throughout the course of the beta.

The number in the first column refers to the bugzilla number under https://bugzilla.redhat.com/

413981 -- Python client does not exit with clean stack trace in some cases - Know issue with some versions of Python
404701 -- At present messages can be published at a rate higher than that by which they are consumed. This results in the queue growing and consuming more memory. (A queue can flush messages to disk when it reaches a certain size; this requires the limits to be set when the queue is created).This will result in the client connection being closed.
408741, 408741 -- As the python and java clients do not support wraparound of the command id sequence a session that sends more than 2^(32-1) / 2 messages will lose its ability to synchronise correctly.
412011 -- session resume not operational in beta 1

[edit]

Notes

It has been noted by a user that the defualts are not set to use the async journal or enbale any of the performnace optimization - this will be updated for the next update to be the defualts.
When using async journal, the available file handles for the process needs to be configured accordingly for number of queues and sizes (2*f)*q where f is the number of files and q is the number of persistent queues. Set max files with the relavant Linux commands.
QLogic PCI 2360/2 Fibre channel controllers driver has a know issue (IO lock out) with AIO, the journal and realtime kernel.
In the c++ API, a session is not safe for concurrent use - related to 410551

[edit]