Setting up Broker Instances

This document contains information on setting up the Broker under UNIX. It assumes that you have successfully installed EntireX using the Software AG Installer. It covers the following topics:


Startup Daemon 'etbsrv'

This daemon runs in the background for administering broker instances. It is installed as etbsrv in the directory <Installation_Dir>/bin.

Start of instruction setTo start the daemon

  • Enter the following command:

    - <Installation_Dir>/EntireX/bin/sagexx start

    This ensures that etbsrv is always running and ready to receive start/stop commands. Note that the startup script sagexx sources the SAG environment file EntireX/INSTALL/exxenv.

Start of instruction setTo stop the daemon

  • Enter the following command:

    - <Installation_Dir>/EntireX/bin/sagexx stop

It is also registered to startup at boot time, therefore the installation generates additional scripts in a location that depends on the operating system:

Operating System Location Note
Solaris, Linux /etc/init.d Recent Linux versions use systemd instead of init scripts.
AIX /etc  
HP-UX /sbin/init.d  

Starting or Restarting the Administration Service

The Administration Service is started or stopped by the broker startup daemon etbsrv.

When the broker has been started successfully, the Administration Service waits for messages from other started brokers. This wait period lasts around 90 seconds.

After this wait period, all brokers are started that have an Autostart value of "yes" that have not already started.

When the Administration Service is restarted, it takes a maximum of 90 seconds until the current system status is displayed correctly.

Setting up the TCP/IP Communication

The recommended way to set up TCP/IP is to define attribute PORT=nnnn and optionally HOST=x.x.x.x|hostname in the TCP-specific section of the broker attribute file.

If no port number is specified, the EntireX Broker kernel uses getservbyname to determine the TCP/IP port on which it will listen for incoming connections. The specified name is the value of BROKER-ID in the attribute file.

An entry for this value must be made in the local machine's /etc/services file, for example:

etbnnn yyyyy/tcp # local host
where etbnnn is the BROKER-ID and
  yyyyy is the intended port number.

This is the same place that local broker stubs will obtain the port information. If getservbyname fails, the default port number 1971 is used. This is the same default port number that the stubs use.

Starting and Stopping the Default Broker

If check box Turn on Autostart for default EntireX Broker is checked during installation, the default broker ETB001 is started.

Start of instruction setTo start the default broker

  • Enter command:

    <Installation_Dir>/EntireX/bin/defaultbroker start

Start of instruction setTo stop the default broker

  • Enter command:

    <Installation_Dir>/EntireX/bin/defaultbroker stop

Note:
Both commands require that you source the EntireX environment file <Installation_Dir>/EntireX/INSTALL/exxenv[.csh].

Running Broker with SSL/TLS Transport

The Broker can use Secure Sockets Layer/Transport Layer Security (SSL/TLS) as the transport medium. The term "SSL" in this section refers to both SSL and TLS. RPC-based clients and servers as well as ACI clients and servers are always SSL clients. The broker is always the SSL server. For an introduction see SSL/TLS and Certificates with EntireX in the Platform-independent Administration documentation.

Before starting the Broker, it must be configured to correctly use SSL as a transport mechanism:

Start of instruction setTo set up SSL

  1. To operate with SSL, certificates need to be provided and maintained. Depending on the platform, Software AG provides default certificates, but we strongly recommend that you create your own. See SSL/TLS Sample Certificates Delivered with EntireX in the EntireX Security documentation.

  2. Modify broker-specific attributes.

    Append "-SSL" to the TRANSPORT attribute. For example:

    DEFAULTS = BROKER
    TRANSPORT = TCP-SSL

    See also TRANSPORT.

  3. Set the SSL attributes, for example:

    DEFAULTS = SSL
     KEY-STORE = /<Install_Dir>/EntireX/etc/ExxAppCert.pem
     KEY-PASSWD-ENCRYPTED = MyAppKey
     KEY-FILE  = /<Install_Dir>/EntireX/etc/ExxAppKey.pem
     VERIFY-CLIENT = N
     PORT=1958

    where 1958 is the default but can be changed to any port number.

    See also SSL/TLS-specific Attributes and SSL/TLS and Certificates with EntireX.

  4. Make sure the SSL clients connecting to the broker are prepared for SSL connections as well. See Using SSL/TLS with EntireX Components.

Starting and Stopping an Additional Broker

A default broker is always created during installation. This broker is started automatically by default.

  1. Create a subdirectory called ETBnnn under $EXXDIR/etb if it does not yet exist, place the attribute file under ETBnnn and name it etbfile.

    Example:

    cd $EXXDIR/etb
    mkdir ETB002
    cp /tmp/your attribute file ETB002/etbfile
  2. The broker can be started by executing shell script etbstart in the /<Install_Dir>/EntireX/bin directory, using the syntax:

    etbstart ETBnnn

    where ETBnnn is the assigned Broker ID (for example ETB001).

  3. The broker can be stopped by executing the etbcmd utility in the /<Install_Dir>/EntireX/bin directory using the syntax:

    etbcmd -bbroker-id -dBROKER -cSHUTDOWN

    Optional: The broker can also be shut down in any of the following ways:

    • etbcmd -blocalhost:port -dBROKER -cSHUTDOWN
    • etbcmd -bipaddress:port -dBROKER -cSHUTDOWN
    • etbcmd -bmachinename:port -dBROKER -cSHUTDOWN

    The port number is needed only when the broker is not running on the standard port.

    See also Broker Shutdown Statistics and Setting up TCP/IP Transport.

Note:
The information given here is independent of hardware type and platform.

Uniqueness Test for Broker ID

To guarantee that a broker ID is unique on one machine, a named semaphore is created at initialization. If this semaphore already exists for this broker ID, initialization is terminated with message ETBE0168, "This instance of broker already running". If as a result of an abnormal broker termination this semaphore cannot be deleted completely, you can force a restart of the Broker with Broker attribute FORCE=YES.

Tracing EntireX Broker

This section covers the following topics:

Broker TRACE-LEVEL Attribute

The Broker TRACE-LEVEL attribute determines the level of tracing to be performed while Broker is running. The Broker has a master TRACE-LEVEL specified in the Broker section of the attribute file as well as several individual TRACE-LEVEL settings that are specified in the following sections of the attribute file.

Individual Settings Specified in Attribute File Section Note
Master trace level DEFAULTS=BROKER 1,2
Persistent store trace level DEFAULTS=ADABAS | CTREE | DIV 1
Conversion trace level DEFAULTS=SERVICE; Trace option of the service-specific broker attribute CONVERSION.  
Security trace level DEFAULTS=SECURITY 1
Transport trace level DEFAULTS=TCP | SSL 1
Application Monitoring trace level DEFAULTS=APPLICATION-MONITORING  

Notes:

  1. For temporary changes to the master or individual TRACE-LEVEL without restarting the Broker, use the Broker command-line utility etbcmd.
  2. For temporary changes to the master TRACE-LEVEL without restarting the broker, use Command Central. See Changing the Trace Level Temporarily.

Attribute File Trace Setting

Trace Level Description
0 No tracing. Default value.
1 Traces incoming requests, outgoing replies, and resource usage.
2 All of Trace Level 1, plus all main routines executed.
3 All of Trace Level 2, plus all routines executed.
4 All of Trace Level 3, plus Broker ACI control block displays.

Note:
Trace levels 2 and above should be used only when requested by Software AG support.

Deferred Tracing

It is not always convenient to run with TRACE-LEVEL defined, especially when higher trace levels are involved. Deferred tracing is triggered when a specific condition occurs, such as an ACI response code or a broker subtask abend. Such conditions cause the contents of the trace buffer to be written, showing trace information leading up the specified event. If the specified event does not occur, the Broker trace will contain only startup and shutdown information (equivalent to TRACE-LEVEL=0). Operating the trace in this mode requires the following additional attributes in the broker section of the attribute file. Values for TRBUFNUM and TRAP-ERROR are only examples.

Attribute Value Description
TRBUFNUM 3 Specifies the deferred trace buffer size = 3 * 64 K.
TRMODE WRAP Indicates trace is not written until an event occurs.
TRAP-ERROR 322 Assigns the event ACI response code 00780322 "PSI: UPDATE failed".

Trace File Handling

Attributes MAX-TRACE-FILES and TRACE-FILE-SIZE are used to avoid a constantly growing ETB.LOG file. The trace is written to file ETB.LOG until TRACE-FILE-SIZE has been reached and a new file is opened. The number of files defined in MAX-TRACE-FILES is kept in addition to the current ETB.LOG file.

Example: If you define MAX-TRACE-FILES=9 and TRACE-FILE-SIZE=100M, the current ETB.LOG will be closed after 100 MB have been written. A maximum of nine backup files plus the current ETB.LOG file are kept.

Protecting a Broker against Denial-of-Service Attacks

An optional feature of EntireX Broker is available to protect a broker running with SECURITY=YES against denial-of-service attacks. An application that is running with invalid user credentials will get a security response code. However, if the process is doing this in a processing loop, the whole system could be affected. If PARTICIPANT-BLACKLIST is set to YES, EntireX Broker maintains a blacklist to handle such "attacks". If an application causes ten consecutive security class error codes within 30 seconds, the blacklist handler puts the participant on the blacklist. All subsequent requests from this participant are blocked until the BLACKLIST-PENALTY-TIME has elapsed.

Server Shutdown Use Case

Here is a scenario illustrating another use of this feature that is not security-related.

An RPC server is to be shut down immediately, using Broker Command and Information Services (CIS), and has no active request in the broker. The shutdown results in the LOGOFF of the server. The next request that the server receives will probably result in message 00020002 "User does not exist", which will cause the server to reinitialize itself. It was not possible to inform the server that shutdown was meant to be performed.

With the blacklist, this is now possible. As long as the blacklist is not switched off, when a server is shut down immediately using CIS and when there is no active request in the broker, a marker is set in the blacklist. When the next request is received, this marker results in message 00100050 "Shutdown IMMED required", which means that the server is always informed of the shutdown.