Figure 36.1 - Three RCONSOLE connection options.
There are many things that should be regularly monitored on a NetWare 4 server. These include memory utilization, communications, and NDS activities. In addition, it is important to be able to monitor user connections and server configuration.
The server console is where the server is managed. It is important to remember that when you are operating on the server, you are interacting directly with NetWare. Server console commands form a limited command set available only from the console. This chapter gives you some familiarity with the more important console commands and utilities.
Most servers are IBM PC's or clones typically equipped with large hard drives and 16 MB or more of RAM. A PC has its own monitor and keyboard. The server's monitor and keyboard are referred to as the server console.
The server console displays a colon (:) as a prompt. The rationale behind using the colon is to remind you that you are operating from the server's console rather than from some other operating system, such as DOS.
DOS function keys (such as F3) are not used on the server's console. The F1 key is used to obtain online help. Table 36.1 summarizes the keys that are available to assist you when entering commands at the server console.
Server Console Command Keys
Command Key Description
Up Arrow The up arrow is used to retrieve the last 100 commands that have been issued at the server console. When more than 100 commands have been issued, only the most recent 100 commands are saved. The up arrow moves through the last 100 commands in the order of the most recent command to the oldest command. Down Arrow The down arrow enables you move down the list of commands. The down arrow moves through the last 100 commands in the order of oldest command to most recent command. Right Arrow The right arrow enables you to move horizontally to the right when you are editing the command currently on the command line. Characters are not erased while using the right arrow. Left Arrow The left arrow enables you to move horizontally to the left when you are editing the command currently on the command line. Characters are not erased while using the left arrow. Home The Home key moves you to the beginning of the command currently on the command line. End The End key moves you to the end of the command currently on the command line. Backspace The Backspace key is used to move horizontally to the left of the command currently on the command line while erasing characters. Del The Del key deletes the character currently above the cursor. Esc The Esc key clears the command line, canceling any command you have typed. Enter The Enter key executes the command currently on the command line.
You can create server batch files that execute a complex series of commands. On the server console, any ASCII text file that ends with the extension NCF is treated like a DOS batch file. You can place any valid server console command, or use the LOAD command to load any NLM in such server batch files.These files should be placed in the SYS:SYSTEM directory to be accessible from the server console immediately. If you place them in another directory, either include the path when invoking the batch file or add the directory to the server's search path using the SEARCH ADD volume:directory command.
Please note that if you issue the SECURE CONSOLE command, you can only load NCF files and NLMs from the SYS:SYSTEM directory.
If you are running RCONSOLE on your workstation(151)accessing a server console(151)while another network Administrator is accessing the same server console, your keystrokes may interfere with the keystrokes issued by the other user. This effect is the same as if you were both typing on the same keyboard. If you are both in different server utility screens, your keystrokes do not conflict, but if you are both at the server console prompt, there could be a problem.
Three types of connections are available for remote users. These options are illustrated in figure 36.1 and are labeled 1, 2, and 3. Option 1 is a direct asynchronous connection between a remote computer and the server. Option 2 is a connection across the LAN. Option 3 is a connection where a remote computer takes over a computer on the LAN.
Figure 36.1 - Three RCONSOLE connection options.
LOAD REMOTE
The REMOTE utility can produce an encrypted password for your use. This encrypted password then is placed in a file called LDREMOTE.NCF, which is used to load REMOTE with the encrypted password. The advantage here is that if a user happens to view the password, it is meaningless. The original password must be entered and REMOTE interprets the encrypted password to enable access.
The following is a sample session where REMOTE is used to produce an encrypted password:
EBDB:remote encrypt
Enter a password to encrypt
>
To use this password use the command:
Load REMOTE -E 73261497661191339933D3
Would you like this command written to SYS:SYSTEM\LDREMOTE.NCF? (y/n)
If you answer Yes to the prompt above, a file LDREMOTE.NCF is created and saved in the SYS:SYSTEM directory.
REMOTE has the additional parameters LOCK OUT and UNLOCK. The LOCK OUT parameter locks out additional remote users. The advantage to this is that there are no other users who can interfere with a session on the server console. The UNLOCK parameter re-enables remote access.
To disable other users from accessing the server through RCONSOLE, type the following:
REMOTE LOCK OUT
To re-enable RCONSOLE access, type the following:
REMOTE UNLOCK
Unlike NetWare 4.0x or NetWare 3.x, NetWare 4.1 does not accept the SUPERVISOR password in place of the REMOTE password.
The asynchronous drivers selected are AIO.NLM and AIOCOMX.NLM. These drivers come with NetWare 4.1. The AIO driver is a runtime module used in conjunction with remote routing. It is designed to enable the server to handle asynchronous input/output (I/O) traffic. The AIOCOMX driver is designed for the COM ports on the server. If the server is equipped with other asynchronous hardware, such as a WNIM+ (Wide Area Network Interface Module) adapter, the driver for this board is loaded in place of AIOCOMX.
Finally, the RS232.NLM module is loaded to enable REMOTE.NLM to accept console through the RS-232 port (or other asynchronous ports) on the server. When you load RS232 you are asked for the port number and speed (baud rate) of the modem attached to the server's asynchronous port.
The following is a sample session on the server console configuring the server for a direct asynchronous connection:
EBDB:load remote Loading module REMOTE.NLM NetWare Remote Console Version 4.10e August 24, 1994 Copyright 1994 Novell, Inc. All rights reserved. Enter a password for Remote Console > Remote Console successfully loaded EBDB:load aio Loading module AIO.NLM NetWare 4.0x Runtime Asynchronous I/O Library Version 2.00 February 14, 1994 Copyright 1991-1993, Novell Inc. All rights reserved. EBDB:load aiocomx Loading module AIOCOMX.NLM NetWare 3.x AIO COMX Driver Version 2.00 December 17, 1992 Copyright 1992, Novell Inc. All rights reserved. For AIOCOMX Driver Load Information, Enter 'LOAD AIOCOMX ?' Installed I/O Address 0x3F8 (Interrupt Number 4) as Board 0, Port 0 8250/16450 device detected: Driver rated at 2400bps EBDB:load RS232 Loading module RS232.NLM NetWare Remote Console RS232 Driver Version 4.10 August 24, 1994 Copyright 1994 Novell, Inc. All rights reserved. Available COM ports: #1 - hardware type:1 board:0 port:0 Please choose a COM port COM Port: 1 Please enter the baud rate (2400 - 38400) Baud Rate: 9600
The RS232.NLM has these additional options:
LOAD RS232 [com port] [modem speed] [N] [C]
The com port can be COM1 or COM2. On some computers you also can specify COM3 or COM4. The modem speed is from 2400 to 38400 bits per second. The letter N indicates you are using a null modem cable. This is a cable used to directly attach a dumb terminal or another computer to a COM port. The letter C indicates that you want to have the server call back a preselected number. The authorized phone numbers should be stored in an ASCII text file CALLBACK.LST you create and save in the SYS:SYSTEM directory. This file contains phone numbers. When the remote computer dials up, the server records its phone number and hangs up. If the phone number is the same as a phone number in the list, the server dials back this phone number and establishes an RCONSOLE session.
The following is a sample session on the server console configuring the server for RCONSOLE access across the LAN: EBDB:load remote
Loading module REMOTE.NLM NetWare Remote Console Version 4.10e August 24, 1994 Copyright 1994 Novell, Inc. All rights reserved. Enter a password for Remote Console > Remote Console successfully loaded EBDB:load rspx Loading module RSPX.NLM NetWare Remote Console SPX Driver Version 4.10a August 24, 1994 Copyright 1994 Novell, Inc. All rights reserved.
The communications software must be of the type where you connect two computers over a phone line, and the client computer takes over the host computer. Examples of this type of software include the following:
Communications software such as the preceding transmits only keystrokes from the client to the host, and only screen updates from the host back to the client. The host computer runs all applications. This communications setup is very efficient and enables you to access programs on the LAN as well as run RCONSOLE. Once you run RCONSOLE on the host computer, you have access to the server console (provided you know the REMOTE.NLM password).
RCONSOLE
The RCONSOLE Connection Type menu appears.
RCONSOLE might not function correctly if you run it under a Windows DOS session. In most cases this is not a problem. On some machines there can be conflicts, causing the session to terminate abruptly. When running RCONSOLE from Windows, a warning message appears telling you that there can be problems. You can proceed and test access on your computer.
If you selected Asynchronous, you see the Asynchronous Options menu. Select Configuration to configure the phone number and characteristics of your modem. The Current Modem Configuration dialog box appears as in figure 36.2.
Figure 36.2 -The Current Modem Configuration dialog box.
Table 36.2 summarizes the options in this dialog box.
RCONSOLE Current Modem Configuration Options
Option Description
Select Device Choices include Hayes Compatible Modem or Null Modem. If you are using a modem, select Hayes Compatible Modem. If your connection is direct using a null modem cable, select this option. (If your modem is not Hayes compatible, you may want to check your budget and consider making a purchase.) Select com port Select the COM port to which your modem is connected. Select baud rate Select the desired modem speed, from 2400 bps (bits per second) to a maximum of 38,400 bps. Modem Reset Command The command string used to reset your modem. Modem Initialization Command The command string used to initialize your modem. Modem Reconnect Command The command string used to reconnect your modem. Modem Dial Command The command string used to cause your modem to dial. Modem Hangup Command The command string used to cause your modem to hang up. User Connection ID An identifying name used by the server to track who you are. Call Back Number The number for the server to call back. This assumes that you have enabled the callback option (see RS232.NLM options).
Figure 36.3 -RCONSOLE Asynchronous Options--Location/phone Number list.
When you finish entering names in the list, from the Location/Phone Number list select the target server and press Enter. RCONSOLE dials the location and creates a connection.
If the desired server does not appear on the list, the proper REMOTE and other NLMs might not be loaded on that server. You can go to the server (if available) and load these NLMs. When you return to your workstation, press Esc and answer No to the question, "Exit Remote Console?" When you return to the Available Servers list, the server checks the LAN for additional servers running REMOTE, refreshing the list.
RCONSOLE Available Options Menu Options
Select A Screen To View Brings up a list of screens used by the various NLMs you have loaded on your server. You then can select which screen to view and press Enter to jump to that screen. Directory Scan Enables you to view a list of files and directories on the server. This includes a view of the server's DOS partition and any floppy disks that may be in its floppy drive(s). Transfer Files To Server Enables you to transfer files to the server. Invoke Operating Takes you to the DOS shell where you can perform any DOS command. System Shell When you type EXIT you are returned to the RCONSOLE session. This is especially useful when transferring files or getting a directory listing. End Remote Session Exits the server console. You also can press Alt+F2 to exit the With Server server console. Resume Remote Session Returns you to the server console, closing the With Server RCONSOLE menu. Workstation Address Gives you the workstation network and node address (useful for viewing connection information). You also can press Alt+F5. Configure Keystroke Enables you to control when your keystrokes appear on the server Buffering console screen. This is useful when there is a process running on the server or a series of error messages that prevent you from seeing your own keystrokes.
The options for Configure Keystroke Buffering are as follows:
You can view directly the NDS files using RCONSOLE. From the RCONSOLE menu, select Directory Scan. When you are asked for a directory name, enter SYS:_NETWARE. When you press Enter, you see the names, dates, and sizes of the NDS database files physically stored on this server.
RCONSOLE has the following keystrokes available (some of which have been mentioned already):
Keystroke Function
Alt+F1 Brings up the RCONSOLE menu. Alt+F2 Exits the server console. Alt+F3 Toggles to the next server console screen. If you have MONITOR.NLM and INSTALL.NLM loaded, for example, repeatedly pressing Alt+F3 toggles from the console prompt to MONITOR,then to INSTALL, and then back to the console prompt. Alt+F4 Toggles to the previous server console screen. If you have MONITOR.NLM and INSTALL.NLM loaded, for example, repeatedly pressing Alt+F4 toggles from the console prompt to INSTALL, then to MONITOR, and then back to the console prompt. Alt+F5 Displays the network and node address of your workstation. Alt+F8 Sends buffered keystrokes to the console if Manual keystroke send keyboard buffering is selected. (See Configure Keystroke Buffering from the RCONSOLE menu.) Alt+F9 Sends buffered keystrokes to the console if On demand buffering keyboard buffering option is selected. (See Configure Keystroke Buffering from the RCONSOLE menu.)
LOAD MONITOR
Figure 36.5 -The MONITOR main menu and display screen.
MONITOR can be loaded with two parameters: L and N. The L parameter causes MONITOR to load with the server console keyboard automatically locked. The password for the bindery user SUPERVISOR can be used to unlock the console (this password is the original password assigned to the Admin user, unless changed). The LOAD MONITOR L command can be placed in the AUTOEXEC.NCF file. The N parameter causes MONITOR to load without the screen saver.
If you leave MONITOR running with no keyboard activity for 10 seconds, the General Information display screen expands to that shown in figure 36.6. You can use the Tab key to toggle between normal and expanded sizes. Another 10 minutes of inactivity causes MONITOR to invoke the screen saver, which consists of a series of consecutively lighter squares displayed on the screen.
Figure 36.6 -The expanded MONITOR General Information screen.
Table 36.4 summarizes the fields in the MONITOR General Information Screen. Most values can be adjusted using the SERVMAN utility or by issuing a SET command from the server console. These commands are covered in the next chapter.
Fields in the MONITOR General Information Screen
Field Description
Server up time The amount of time in Hours:Minutes:Seconds the server has been up. Utilization Percentage of CPU utilization of the server. Original cache The number of cache buffers (4 KB each) available when the server buffers first booted, before any NLMs were loaded. Total cache buffers The number of buffers remaining after the various NLMs and memory allocations have taken place. This is the amount of RAM available to cache files requested by users. The ratio of Total cache buffers to Original cache buffers should be 50 percent or greater. If not, you may want to load fewer NLMs or add more memory. If the ratio goes below 20 percent, the operation of the server is severely affected. Dirty cache buffers The number of buffers with data waiting to be written to the hard drive(s). If the ratio of Dirty cache buffers to Total cache buffers is greater than 75 percent, there is cause for concern. This means that there is a bottleneck getting data written to the hard drive. You can set a parameter: set Maximum disk cache writes to a higher value, or look at improving your disk hardware. Current disk requests The number of requests for data to or from the hard drives. Packet receive buffers The number of buffers reserved for processing requests between the server and client. The server continues to allocate packet receive buffers above the minimum number you set on an as-needed basis. If users report a slowdown in response, check if the server has allocated its maximum. If this is the case you can increase this value by issuing the Set Maximum Packet Receive Buffers command. Directory cache The number of buffers allocated for directory caching. The target for buffers data cached in this buffer space is the Directory Entry Table (DET), which uses another portion of RAM. When users need to access the DET frequently, the server allocates additional directory cache buffers to meet the needs. Maximum service The maximum number of service processes (task handlers) currently processes allocated to fill user requests. Each service process takes care of a file system request (such as reading files from or writing files to the server). When the task is completed, the service process is put to sleep. Current service The number of task handlers currently activated. This number processes represents the maximum currently made available by the server. The number of service processes actually working may be lower. If you notice that the number of current service processes equals the maximum service processes you should increase the maximum using the Set Maximum Service Processes parameter. Maximum licensed The number of user or other connections which are allowed by your connection version of NetWare. Current licensed The number of connections currently in use which count against your connections NetWare license. Open files The total number of files open on the server.
The following sections describe some of the MONITOR menu options. Because the options are too numerous for coverage in this chapter, only the items which have the most critical impact on your system are covered.
Figure 36.7 -The Active Connections list.
A connection is a logical designation used by the server to track a communications session. When you first load the VLM shell on a DOS workstation, for example, you see an initial message, "Attached to Server X," where X is the name of the server. This attachment is assigned a connection number. If you look at the Active Connections list at this point, you see the connection number followed by the name NOT-LOGGED-IN. Once you login, your login name replaces NOT-LOGGED-IN.
If you press Enter after selecting a connection on the Active Connections list, you see the Connection Information screen as shown in figure 36.8.
Figure 36.8 -The Connection Information screen
As you can see from figure 36.8, the Connection Information screen gives you information on the activities of this particular connection. You can see which files are currently open. This information is useful when the user is running a process that terminates and locks up the workstation for no apparent reason. By monitoring this user's connection information you can see which files are open at the time of the lockup.
Other information recorded includes the following:
When you press Enter on any open file, you can see its file or record lock status. If you press Del after selecting a connection from the Active Connections list, you can clear the connection. This causes any files opened by this connection to close, and logs the user out of this server.
Figure 36.9 -The System Disk Drives list.
Hard disk or CD-ROM drives are assigned device numbers. Following the device number is a description of the device giving you manufacturer and model information. In the figure you see device 0, "MEDIAVISCDR-H93MV," and device 1, "QUANTUM LP240A GM240A01X." Device 0 is a MediaVision CD-ROM drive, model CDR-H93MV. Device 1 is a Quantum IDE hard drive, model ProDrive LPS 240AT. The number following each device in parentheses is an internally generated device identifier. It consists of a manufacturer driver ID number, card number, controller number, and drive number. For example, the ID assigned to the Future Domain driver (a SCSI adapter used to drive the MediaVision CD-ROM in this example), is 81.
When you select a device in the System Disk Drives list, additional information appears. You can press the Tab key to expand the top screen. The resulting Drive Status screen is shown in figure 36.10.
Figure 36.10 -The Drive Status screen.
At the top of the Drive Status screen are the driver description and the card, controller, and drive numbers. The bottom screen is discussed later. Table 36.5 summarizes the contents of the top screen.
Fields in the Drive Status Screen
Field Description
Driver The NLM loaded to drive this device. Disk size The size of the device. Partitions The number of partitions on this device. Partitions can include DOS, NetWare, OS/2, UNIX, Windows NT, or other operating system partitions. Mirror status Indicates if this device is part of a mirrored set. Hot Fix status If this field shows Normal, then Hot Fix is enabled and functioning normally. Hot fix is the process by which bad blocks are marked bad and data is diverted into the hot fix area. Partition blocks The number of disk blocks available in the NetWare partition. The block size used in this screen is 4,096 (4 KB). Data blocks This represents the number of data blocks out of the total number of partition blocks. Redirection blocks The number of blocks reserved for hot fix, out of the total number of partition blocks. Redirected blocks This is a critical number. This is the total number of blocks, out of the redirectio blocks, used because bad blocks were detected. This number equals the number of bad blocks found on the NetWare partition of this drive. Reserved blocks The number of redirection blocks reserved for the hot fix process.
At the bottom of the Drive Status screen is a separate list. The most interesting aspect of this list comes into play when the drive is a CD-ROM drive, or has removable disk cartridges. In this case, the Drive Status list at the bottom changes, adding two additional parameters: Removable Drive Mount Status and Removable Drive Lock Status. If you want to dismount a CD, for example, you change the Removable Drive Lock Status to Not Locked. You then change the Removable Drive Mount Status to Dismount. This causes the CD to dismount. This technique is not recommended, however, and a better technique is to use the following command:
CD DISMOUNT
volume is the name of the volume to dismount. This command assumed you have CDROM .NLM loaded. For more information on dismounting CDs, see Chapter 37, "Making Configuration Changes."
A frame type is a particular format used to send and receive data from this network adapter. Certain protocols are geared towards certain frame types. The TCP/IP protocol suite, for example, is oriented towards the Ethernet_II, Ethernet_SNAP, or Token-Ring_SNAP frame types. See Chapter 3, "LANs and WANs," for more details.
When you press Enter after selecting a particular LAN driver, a statistics screen appears, as shown in figure 36.11.
Figure 36.11 -The LAN/WAN Driver Statistic screen.
A detailed discussion of every parameter is beyond the scope of this book. Table 36.6 summarizes some of the general statistics you should monitor.
LAN/WAN Statistics To Monitor
Statistics Description
No ECB available count An ECB is an Event Control Block. This is an engineer's term for a packet receive buffer. If you see this statistic increase, check if the packet receive buffers statistic on the main MONITOR screen has reached its maximum. If this is the case, increase the maximum using the SET MAXIMUM PACKET RECEIVE BUFFERS command. Receive Packet Too Big This indicates the number of times a frame of data was received which was too large for the packet receive buffer to handle. If this number is large in proportion to the total (more than 1 or 2 percent of the total), increase the size of the packet receive buffers using the SET MAXIMUM PHYSICAL RECEIVE PACKET SIZE command. Receive Packet Too Small Any of these errors can point to a bad network adapter. A network that is Miscellaneous Errors going bad sends out frames of data which are too big, too small, or which have Checksum Errors internal errors. This also could be a sign of a cable which is intermittently shorting out, or a loose connector. Adapter Queue Depth If this statistic increases it is a sign that the network adapter in the server is having difficulty getting data onto the LAN. A possible cause is a busy LAN (perhaps too busy). Look at the total number of packets sent and received, and check if the consensus among users is that the network is slow. You can segment the network by breaking up the physical cabling and placing routers between segments. This decreases the traffic on each segment. Another possible cause is that the adapter on the server is insufficient for its job. You may consider replacing it with an adapter with internal buffering or with better throughput. If the adapter is a 16 bit ISA bus adapter, for example, consider replacing it with a 32 bit EISA adapter (if this is appropriate for your system). Excess Collisions A collision occurs when two or more network adapters transmit on an Ethernet network at the same time. Excess collisions, on the other hand, indicate that the network is either too busy and needs to be segmented, or that you have a bad network adapter which is sending out spurious transmissions. Abort Delimiter Specific to token ring, an excessive number of this error indicates that the active monitor in a token ring environment had to reinitialize the ring. This can indicate a bad cable, faulty MAU, or a network adapter going on and off line frequently. Frame Copied Error In a token ring environment this indicates the number of times the destination station received a transmission which another station had already copied. This can indicate physical wiring problems(151)a line hit by EMI (electromagnetic interference), for example, or a duplicate node address.
Figure 36.12 -The System module information list.
Figure 36.13 -The File open/lock activity file status screen.
This is especially useful when you have problem applications that lock up workstations, or where users are experiencing problems deleting, modifying, or changing attributes of files. If a file is considered open, it cannot be deleted or modified in any way until closed.
The statistics screen shows the Use count and the Open count parameters. These parameters are the number of connections that have this file open and in use. The Open for read and Open for write parameters display the number of connections that have this file open for either reading or writing. When more than one connection has a file open for write, a possible conflict can arise unless the application in use is capable of controlling multiuser write access.
The Deny read and Deny write parameters indicate how many connections want to prevent others from reading from or writing to the file. The Open status parameter indicates the current status of the file when opened. At the bottom of the screen you see a list of connections that have the file open. If you need to close down this file for any reason, you can go to the Connection information screen and clear those connections with the file open. Clearing connections also logs that user out.
Figure 36.14 -The Cache Utilization Statistics screen.
Some terms you see on this screen include the following:
The statistic to watch on this screen is Long term cache hits. This figure should be above 90 percent. If it falls below 80 percent, network users notice a slowdown in network performance. If the situation continues to deteriorate, you should add RAM to the server.
When the LRU sitting time is large, the oldest allocated cache block is idle and has not been re-used. This is an indication that either your server is not busy, or you have ample RAM.
For information on the other statistics, use the F1 key for help.
Figure 36.15 -Processor utilization--Interrupt 0 through Interrupt 15 processes..
In the figure you can see activity for interrupts 0, 5, 11, and 14. In this example, a workstation copies files from the CD-ROM drive to the server's hard drive. Interrupt 0 is the system clock. Interrupt 5 is used by the Intel Ethernet adapter on the server. Interrupt 11 is used by the SCSI controller to which the CD-ROM drive is attached. Interrupt 14 is used by the server's hard drive controller.
It is important to note that the Load Factor you can see in figure 36.15 is a percentage of the current CPU utilization. This statistic is visible from the main MONITOR General Information screen next to the Utilization parameter.
Figure 36.16 -The Tracked Resources and Server Memory Statistics screens.
The Tracked Resources list displays all the system resources that are tracked by this option. When you press Enter, the Resource Tags list appears showing you which NLMs are making use of this resource. A tag is a name assigned when an application (an NLM) requests, and is granted, a resource by the operating system.
At the top of the screen you can see a breakdown of server memory in the Server Memory Statistics screen. The Cache buffers parameter is of the most interest to network Administrators. This represents the percentage of memory available for user file caching. If this figure is above 50 percent your server is in good shape. If the figure is between 30 and 50 percent the amount of RAM is marginal to adequate, with little or no reserves. If the figure drops below 30 percent, you should add more RAM to the server. Usually your network users report a slowdown in performance by the time this figure drops below 30 percent.
Figure 36.17 -The Allocated Memory For All Modules and System Modules screens.
In the figure, a little more than 2.5 MB is in use from the allocated memory pool. You can go through the System Modules list one NLM at a time and press Enter to get an update on allocated memory usage for that particular module.
Some modules may allocate large blocks of memory for a short period of time and then drop down in use. In this case, the Percent free parameter represents memory which can be used in the user file caching memory pool. The garbage collection process takes unused memory (in 4 KB pages) and returns it to the main memory pool (the user file caching memory pool). You can force this process to take place for a particular module by proceeding as follows:
Garbage collection can only release memory in 4 KB pages. If the module only has allocated one cache page, and any percentage of that one page is in use, then garbage collection cannot occur for this module. For example, the Future Domain CAM SCSI Disk Driver has allocated one 4 KB cache page. The Percent in use parameter shows 1 percent. The Percent free parameter shows 99 percent. When you press F3 you notice that there is no change. The reason is that garbage collection does not release memory if it is less than 4 KB.
Pressing the F5 key when the Allocated Memory For All Modules screen is expanded performs garbage collection for all modules. Otherwise, the garbage collection process takes place at regular intervals. See Chapter 38, "Fine Tuning the Server," for more information on using the SET commands to adjust parameters such as garbage collection.
Figure 36.18 -The Scheduling Information screen.
Scheduling Information Statistics
Statistics Description
Process Name The name of the process being monitored. Sch Delay The delay factor used when scheduling a process. The default is 0, which means no delay. You can press the plus key (+) at the right side of the console keyboard to introduce a delay in a process. Time The time the CPU spent servicing this process. Count The number of times this process was serviced in this sampling interval. Load Percentage of CPU time spent on this process. Interrupts The combined total time, count, and load spent servicing interrupt processes. Idle Loop The time the CPU spent not doing anything at all. This represents the margin of CPU power available to you. Subtract this value from 100 percent to arrive at the total time the server's CPU spent working. Work Other miscellaneous processes. Total sample time Total CPU time measured during this sample. Histogram overhead The overhead associated with the Scheduling information time process itself. Adjusted sample time The amount of time minus the Histogram overhead time.
You might find that certain processes are "hogging the CPU." This means that they are using more CPU time than is desirable. The more time a process spends on the CPU, the less time is available for other processes. You might want to introduce a delay to a process deliberately. To do so, select the process and press the plus key (+) at the far right side of the console keyboard (use the numeric keypad plus key). The Sch Delay parameter increments. Each time you increase the Sch Delay, you cause the CPU to schedule this process less often. A value of 2, for example, causes the CPU to schedule this process 1/2 as often as a process with a delay of 0 (there is not an option of 1/1). A value of 10 causes the process to be scheduled 1/10 as often as a process with a delay of 0.
LOAD SCHDELAY [process = value]This causes process to be delayed by value. To make scheduling delays permanent, add the SCHDELAY command to the AUTOEXEC.NCF file. To delay the REMOTE process by a factor of 2, for example, issue the following command:
EBDB:load schdelay remote = 2
Loading module SCHDELAY.NLM
NetWare 386 Scheduling Delay Control
Version 1.02 February 1, 1993
Copyright 1993 Novell, Inc. All rights reserved.
Debug symbol information for SCHDELAY.NLM loaded
Setting "Remote" Scheduling Delay to 2.
EBDB:
To determine process names, load SCHDELAY with no parameters.