Read this chapter if one of the following statements is applicable:
This is the first installation of NQE at your site.
You are upgrading and plan to change your NQE configuration.
Otherwise, skip to the appropriate set of installation procedures.
The following sections describe configuration decisions you must make before you start the installation process, including the following:
Planning server type(s) and clients
Planning server configuration
Planning an NQE cluster configuration
Setting port numbers
An NQE cluster can contain the following components defined by the NQE_DEFAULT_COMPLIST variable in the nqeinfo(5) file:
| NQS | Network Queuing System | |
| NLB | Network Load Balancer | |
| COLLECTOR | NLB collector | |
| NQEDB | NQE database | |
| MONITOR | NQE database monitor | |
| SCHEDULER | NQS scheduler | |
| LWS | Lightweight server |
Beginning with the NQE 3.3 release, the default component list consists of the following components: NQS, NLB, and COLLECTOR.
NQE clients contain software that allows users to submit, monitor, and control requests by using either the NQE graphical user interface (GUI) or the command-line interface. From clients, users may also monitor request status, delete or signal requests, monitor machine load, and receive request output.
NQE clients also contain the NQE Web example, which provides a sample World Wide Web (WWW) interface to selected NQE functions. After it is installed, you can access the WWW interface through Web clients such as Mosaic or Netscape. This means that IBM PCs, Macintosh systems, workstations, and superservers can use NQE services. Through the WWW interface, you can submit a batch request (from a file or entered interactively), obtain status on your requests, delete your requests, and signal your requests. You also can view and save output files produced. Hypertext-based online help is provided.
If you are not sure where you want to run NQE components and clients, you should read NQE Administration, publication SG-2150, which provides detailed descriptions of NQE components and contains an appendix that introduces the concepts of simple networks, load-balanced networks, and the default NQE configuration.
After you have determined where in your NQE cluster you will run NQE components, clients, license servers, and other installation options, you must gather information about the hosts on which they will run. This information is used when you contact your NQE customer support representative to obtain your permanent NQE 3.3 license and during installation of NQE.
![]() | Note: If you will run the NQE database, you must select only one NQE database node (MSQL_SERVER) to run the NQE database, scheduler, and monitor. This is done by including NQEDB, SCHEDULER, and MONITOR in the list of configured NQE components specified for the NQE_DEFAULT_COMPLIST nqeinfo variable.
Also, you must configure the MSQL_SERVER host and the MSQL_TCP_PORT TCP/IP port for the NQE database mSQL server. These two settings must be the same on every machine (node) in the NQE cluster. |
You must have information about host names, directories, and port numbers for the servers and clients you install. Table 7-1, describes the necessary directories and host names, lists disk space requirements, and identifies the defaults that the configuration procedure uses.
Table 7-1. NQE Configuration Information
Item | Description | Default value |
|---|---|---|
target_root file system | The NQE version maintenance utility, nqemaint(8), creates a directory tree under this directory. The NQE directory is installed in a subdirectory of the target root file system called /craysoft. Do not use a tmpfs (memory-based) file system.
Systems other than UNICOS and UNICOS/mk need the following approximate amount of disk space, depending upon the system. NQE components and clients: approximately 22.7 Mbytes to 33.1 Mbytes of disk space. Client: approximately 9.25 Mbytes to 14.2 Mbytes of disk space.
UNICOS and UNICOS/mk systems need approximately 95 Mbytes of disk space. | /opt on Solaris, UNICOS, and UNICOS/mk systems and normally /usr on all other platforms. On UNICOS and UNICOS/mk systems, to install NQE, the target_root file system must be /opt. |
/tmp file system | All UNICOS and UNICOS/mk systems require 95 Mbytes of disk space. | |
NQE spool directory | All servers must specify the directory in which NQE working files (such as request files, output files, and the NLB database) are placed. For an initial installation, you need approximately .5 Mbytes of disk space, although the amount of required disk space will vary depending upon the type of logging you do and the number of jobs running. | /usr/spool/nqe |
NLB server host name | All NQE nodes must specify the host name of the NQE nodes on which the NLB server runs. To ensure greater availability of the NLB server in the network, you can run redundant servers. | Local host |
NQE database server host name (mSQL server host) | If you will use the NQE scheduler and NQE database, all nodes and clients must specify the host name of the NQE node on which the database server runs. Only one NQE node in the NQE cluster can run the database server, even if you install multiple NQE nodes. | Local host |
NQS_SERVER | All clients must have a default NQS server host to process their requests. This value must be the host name of an NQS server. | Installation host for NQS servers; no default if installing on client |
When planning a cluster configuration, you will need to add mids (machine IDs) for all servers in the cluster to the NQS configuration on each server. Also, any NQS batch queues and pipe queues in addition to the default configuration should be added at installation time. This may be done by adding all mids and queues on the first server you install and then creating a snap file as follows:
% qmgr Qmgr: snap file = /tmp/snapfile |
![]() | Caution: When you copy the snap file into ${NQE_ETC}/nqs_config on other servers, ensure the NQE_ROOT has the correct value for the system. For example, a snap file written on an IRIX system, where the value of NQE_ROOT is /usr, will have the wrong path names if it is read without modification on a UNICOS system, where the value of NQE_ROOT is /opt. |
Modify the snap file as needed so that paths match NQE_ROOT. In the case of pipe queues, the absolute path is not needed. For example, the default nqenlb load-balancing queue could be changed from the following:
create pipe_queue nqenlb priority=63 \ server=(/nqebase/bin/pipeclient CRI_DS) |
to the following:
create pipe_queue nqenlb priority=63 server=(pipeclient CRI_DS) |
Servers with multiple network interfaces must have all interfaces named for the same mid, or a machine ID conflict is likely. If you are letting NQE calculate the mid, after NQE has calculated all the mids on one server, use those mid numbers when adding mids on all other servers. Using the above snap file method to update the configuration on every other server will accomplish this for you. For example:
stone% netstat -i Name Mtu Network Address Ipkts Ierrs Opkts Oerrs np1 16432 hyp/24 stone 45433 0 8303 38 fd0 4352 fddi/2 stone-fddi 1571299 1 836739 0 lo0 65535 loopback localhost 4905 0 4905 0 stone% qmgr Qmgr: add mid stone stone-fddi Qmgr: add mid clay clay-hyp Qmgr: show mid MID PRINCIPAL NAME AGENTS ALIASES -------- -------------- ------ ------- 10638371 stone nqs stone-fddi 10640917 clay nqs clay-hyp clay% qmgr Qmgr: add mid 10638371 stone stone-hyp Qmgr: add mid 10640917 clay clay-hyp |
If a machine ID conflict occurs, the following error message would appear in the NQS log on the server:
02/06/97 14:30:30:6777: prenetserver: 15097: 10485: netdaemon.c: 960: nqs-617: WARNING: pre_server(): Conflicting mids for client : cmid <10638371>, lmid <10638362>. |
In this case, the mid for stone on stone (the client) is 10638371, but the server has 10638362 as the mid for stone and does not recognize stone.
If the server does not have a mid for stone, the following error message appears in the NQS log:
02/06/97 17:21:14:7567: prenetserver: 15097: 7047: netdaemon.c: 936: nqs-614: WARNING: pre_server(): Client host name unknown to local host. |
The TCP/IP port numbers for NLB, FTA, the NQE database (mSQL), and NQS have default values, as shown in Table 7-2. These numbers must be the same for all nodes in the NQE cluster.
You must ensure that no other applications are listening on these ports. To display the ports that are in use, use the netstat -a command.
To determine the port numbers that are configured for the local system, examine the /etc/services file. If you are running NIS or NIS+, use the ypcat command to search the services file for port numbers. For example, to determine if the TCP/IP port number configured for NQS is port 607, use the following command:
% ypcat services | grep 607 nqs 607/tcp # Network Queuing System |
![]() | Note: For UNICOS and UNICOS/mk systems, the FTA port changed from 21 to 605 as of the NQE 3.1 release. |
Table 7-2. NQE Default Port Numbers
Item | Description | Default value |
|---|---|---|
mSQL port | TCP/IP port number for the NQE database (mSQL) server | 603 |
NLB port | TCP/IP port number for NLB | 604 |
FTA port | TCP/IP port number for file transfer | 605 |
NQS port | TCP/IP port number for batch request transfers | 607 |