Managing the Server

This section describes basic setup and configuration of your NetWare server so that it will run efficiently. It explains how to do the following:

Start and stop your server
Enable logins and server console messages
Start your NCP engines and enable packet burst
Access reports on clients
Set specialized server-wide security variables
Configure startup variables
Enable IPX auto-discovery and IPX diagnostics
Configure your IPX and SPX service advertising

Starting NetWare after Booting Your System

You can configure NetWare to come up automatically when you boot the server. If you configure it this way, you can use the nwserverstatus command to check that NetWare really did come up after the reboot.

The system parameter that controls the server starting at boot is nws_start_at_boot. If you cannot start NetWare, you could check nwcm at the HP-UX serverc console command line by typing:

nwcm -v nws_start_at_boot

This tells you if the parameter is set to On or Off. If it is off, type:

nwcm -s nws_start_at_boot=on

to turn it on again.

Bringing Down and Restarting NetWare

Use this procedure to start and stop your NetWare server.

Make sure all of your users are logged out of the server if you shut it down.

Prerequisites

Access to the server console
Superuser permission to shut down a NetWare server

Procedure

Start the NetWare transport using the command:
startnps
Start NetWare services using the command:
startnw
In order to start the NetWare server process, the IPX protocol stack must be active.
There are several ways to check at the HP-UX server console command line whether the stack is active:
- Type ps -ef | grep IPX. If the transport is UP , it should display the two stack daemons npsd (IPX Protocol Stack Daemon) and sapd (IPX Sap Daemon).
- Type ipxinfo. A Resource temporarily unavailable message may appear. This indicates that the stack is not active.
- Type nwsapinfo or statnps. nwsapinfo may display an "unable to attach to shared memory" message if the transport is down.
- statnps -i will display UP or DOWN to indicate the status of the transport.
To start the stack again, type startnps at the command line.

Additional Information

For more information about	Refer to
Starting or stopping the server	"nwcm" in the Command Reference

Tuning Your Server

This section discusses how to limit or increase the HP-UX resources available for NetWare Services. It also describes how to manage engines, configure shared memory size, limit error log file size, and set packet burst limits.

Managing Engines at the Desktop

NetWare Services servers refer to certain processes as NCP engines. NCP engines are often considered the workhorses of NetWare Services servers and are responsible for initial processing of all client requests.

NetWare Services requires a minimum of two engines running at all times. The more work you require of your system, the more engines you will need to start.

Adding too many engines, however, may decrease performance. So generally you need two engines per CPU.

Running more engines requires more RAM. (See "Hardware Requirements" on page vii in "How to Use This Manual.") After the NetWare server has booted, engines can be increased or decreased using the nwengine utility. The number should be increased if the nxinfo utility reports an excessive number of packets dropped because of server activity.

To prevent NetWare from using too many resources, you can limit the number of engines that can be started after boot time.

You can specify the number of NCP engines to start by using "Number of NCP Engines to Start" and "Maximum Number of NCP Engines" at the NetWare® 4.1/9000 server console. To specify how many engines to start automatically when the NetWare server process is started, see "Tuning Your Server at the Desktop" in this chapter.

Configuring Shared Memory

If you don't configure any shared memory or if the default size of 4 Mg (or 4194304 bytes) is configured, NetWare will automatically determine the required shared memory size. The actual size used will be displayed on the console at startup.

Adjusting the Size of Your Error Log

Any errors your server experiences are saved to the error log file, SYS$LOG.ERR. You can change the size of this log depending upon how many messages you want to store in this file. Make sure you monitor this file because if messages come in and the file is full, messages are saved to a backup (SYS$LOG.OLD) and a new file, SYS$LOG.ERR, is created. Only one SYS$LOG.ERR.OLD is maintained.

See "Error Log File Size" in "Tuning Your Server" in this chapter to change the size of your error log file.

Configuring Packet Burst

Packet Burst is a protocol built on top of IPX that speeds multiple-packet NCP reads and writes of files.

This protocol speeds the transfer of NCP data between a workstation and a NetWare server by eliminating the need to sequence and acknowledge each packet.

Packet Burst protocol is more efficient than the one-request/one-response protocol in earlier NetWare versions. With this protocol, the server or workstation can send a whole set (burst) of packets before it requires an acknowledgement.

By allowing multiple packets to be acknowledged, Packet Burst reduces network traffic.

It also monitors dropped packets and retransmits only the missing packets.

A client must negotiate the use of Packet Burst with a server before using it.

See "Enable Packet Burst?" and "Packet Burst Buffer Size" in "Tuning Your Server at the Desktop" in this chapter to configure Packet Burst.

Packet burst can only be configured via the nwcm utility. The parameters are:

burst_mode_buffer_size: size of packt burst packets, default = 24576 burst_mode_clients: maximum number of burst_mode clients that can be supported at the same time. burst_mode_protocol: Enable/disables packet burst

Managing Engines at the Command Line

You can use the nwengine utility at the server console command line to start and stop engines while NetWare Services is running. To change the number of engines, type

nwengine number <Enter>

Replace number with the new number of NCP engines you want running. The nwengine command determines the difference between the total number of NCP engines requested and those already running.

For example, if you have four NCP engines already running and you want to add three more engines, type the following command:

nwengine 7 <Enter>

Three more NCP engines will be started.




	NOTE: This number can be increased only to the maximum number of engines specified by the "Maximum Number of NCP Engines" variable.

To reduce the number of NCP engines from seven to five (thus stopping two NCP engines), type

nwengine 5 <Enter>

You cannot decrease the number of engines below two. This is the minimum number of engines required to run NetWare Services.

Additional Information

For more information about	Refer to
NCP engines	"nwengine" in Utilities Reference
	"NCP_engine" in Concepts

Checking Client Connections Using nxinfo

Use the nxinfo command at the server console command line to get a report on client connections. The report provides two pieces of information important to managing the number of engines you want active:

The maximum number of clients ever active at one time. (If this number matches the number of licensed connections, you may need to buy more licenses.)
The clients now active.

By checking these two figures while your server is performing an average work load, you can determine the number of NCP engines to best manage your workload.

If a large number of packets is dropped because a server message indicates that the server is busy, you may need to add another engine. Adding too many engines, however, may decrease performance.

For more information, see "nxinfo" in Utilities Reference.

Checking Client Activity

Use this procedure at the HP-UX server console to configure the NetWare watchdog (a connection-monitoring feature).

The watchdog sends packets to inactive connections and eventually logs out connections that remain unresponsive. Use the watchdog variables to configure how much watchdog information is recorded or displayed and how long an inactive connection can remain unresponsive before it is logged out.

Watchdogs can only be configured with the nwcm command. The parameters watchdog are listed below.

console_display_watchdog_logouts. This variable controls messages sent to the server's display device when the watchdog logs out a connection. "Yes" means messages are sent to the console. "No" means no messages are sent.

If connections are unexpectedly losing their connection to the NetWare server, set this variable to "Yes" until the problem has been corrected. The default is No.

number_of_watchdog_packets. This variable specifies the number of unanswered watchdog packets the NetWare server sends to a client before it disconnects. Valid values are 5 to 100. The default is 10.

log_watchdog_logouts. This variable controls the creation of log entries when a client is logged out by watchdog. "Yes" means the entry is logged into SYS$LOG.ERR. "No" means the entry is not logged into SYS$LOG.ERR or any other error file. The default is Yes.

delay_before_first_watchdog_packet. Delay before First Watchdog Packet. This variable determines the time (in seconds) the server waits without receiving a request from the workstation before sending out the first watchdog packet to that workstation. The default is 296.

delay_between_watchdog_packets. This variable determines the time (in seconds) between watchdog packets. After the server sends out the first watchdog packet, it waits a specified time before sending out succeeding packets if there is no reply. The default is 60.

For more information on Watchdog, see "Watchdog" in Concepts.