6.2 Creating a Core Dump

Novell might request a core dump when a server experiences a lockup or ABEND and other troubleshooting has failed to resolve the problem. A core dump is a copy of a NetWare server’s memory at the time the server ABENDed. CORE dumps can be analyzed by Novell engineers and are often the key to finding software bugs. However, now that servers typically use large amounts of RAM, creating a core dump is less likely to be a routine troubleshooting procedure. Your abend.log file could be equally valuable for diagnosis.

IMPORTANT:You will be expected to install all necessary patches before Novell requests a core dump. To send a core dump file to Novell, you must first open a support incident, which will be billable. You are not charged until the incident is resolved or closed. If the problem is a Novell bug and no patches were previously available, then there is no charge.

You can start a core dump in these two ways:

6.2.1 Core Dump Prompt Options

Whether you selected the core dump option after an ABEND or are forcing a core dump, you have the following options when responding to the core dump prompts.

Table 6-2 Core Dump Prompt Options

Prompt

Option

Reboot server after coredump?

Y : Reboots server after the core dump

N : Leaves the server in the same state after core dump

Lists available local drives and available free space on each

Lets you choose a local partition, including USB devices, for the dump.

Coredump Type

Full copies all server memory to the specified location.

The size of the full image file will be approximately equal to the total RAM installed in the server.

Full w/o cache copies all server memory except file cache (disk cache) to the specified location.

The size of the image file will be approximately equal to total RAM minus the amount of file cache (disk cache memory).

The cacheless core dump is smaller but it provides as much useful information as a full core dump.

Compress Coredump

Yes takes longer to perform the dump but uses less disk space

No takes less time to perform the dump but uses more disk space

Where should diagnostic core dump be sent?

HDISK sends the file to a local partition. For details, see Sending a Dump to a Local Partition.

Network sends the file to the IP address of the image host (server or workstation)

This greatly decreases the time that the server must remain in a downed state for diagnostics.

For details, see Sending a Dump to a Network Device.

Sending a Dump to a Local Partition

After you start the core dump, you are asked to specify the drive letter and file path that the memory image file will be written to.

The drive can be any writable local device that contains enough storage space. This could be the C: drive or another FAT partition (on a USB storage device, for example).

IMPORTANT:If the device is a USB device, you must format it and plug it in to the server so that it has a drive letter assigned to it before the core dump is initiated.

You can plug in a preformatted USB storage device when the server is halted in the debugger (# prompt). The device is recognized and mounted automatically as a FAT partition. Then, when you initiate the core dump with the .c option, the USB storage device shows up as one of the local drives with the associated drive letter.

You can also create a logical core dump drive using Novell Remote Manager and save the core dumps to this drive just as you would to the local C: drive or other local FAT partitions. For more information, see Creating a Logical Volume for Saving Core Dumps in the Novell Remote Manager for NetWare Administration Guide for OES.

When the image is written to a local hard disk on the server, the default name of the image file is c\:coredump.img.

To write to a different device, substitute the drive letter and filename. For example, d:\coredump.img

After the server is up and running, the image file can be compressed, backed up to tape, or sent to Novell (if you have opened a support incident).

The image file can also be written to a network drive later. This can be done using sys:system\imgcopy.nlm or any other third-party NLM that provides this functionality.

Sending a Dump to a Network Device

  1. Start an image host agent by doing the following:

    On a NetWare server, load imghost.nlm

    or

    From a Windows workstation, run sys:public\imghost.exe.

    When loading the imghost.nlm or running the imghost.exe, you can specify a directory where you want the image to be sent (the directory must already be created) or accept the default paths.

    For example: load imghost sys:coredump

    The default path for imghost.nlm is volume sys:.

    The default path for imghost.exe is the directory that you run the .exe from.

  2. At the “Where should the diagnostics coredump be sent?” prompt, select the Network option.

  3. At the “IP Address of IMGHOST” prompt, enter the IP address of the server or workstation where you started the image host agent.

The image host agent receives the core dump from server and stores it in a directory based on the server name and the file name based on the date (year, month, day, version) of the core dump with an .img extension.

For example, if a file was sent from a server named DASERVER1 on April 3, 2003 to a server running imghost.nlm, the DASERVER1\030403aa.img directory and file would be created on volume sys:.

6.2.2 Sending the Core Dump to Novell

Before sending the memory image to Novell, contact a support representative to open a support incident. You will be assigned to a Support Engineer who will help you analyze the memory image file. The Support Engineer will make arrangements to receive the image either in the regular mail or through the Internet and will advise you on the best media format to use.

6.2.3 Modifying the Defaults of DBNET

The dbnet6.nlm controls the network settings for the sending a core dump to the network. The dbnet6.nlm is loaded by default on the server and it automatically configure itself by scanning for any network boards bound to TCP/IP. It chooses a board that is on the same network as the default router specified by tcpip.nlm. If no default router is present, DBNET chooses the first IP network board found.

The dbnet6.nlm can also be used to display statistics that might assist you in debugging problems with your server. For information on the command available from the System Console, see Using DBNET Console Commands.

After dbnet6.nlm is loaded, you can manually save its settings to a configuration file (dbnet.cfg) in the server’s boot directory (typically c:\nwserver). This procedures is optional. For procedures, see Manually Setting the DBNET Configuration (Optional).

Using DBNET Console Commands

The following is a table of console commands that are available when dbnet6.nlm is loaded:

Table 6-3 Console Commands for dbnet6.nlm

Command

Purpose

dbhelp

Displays DBNET console commands

dbconfig

Displays DBNET configuration

syntax: [save | config | reset]

The Save option saves DBNET's settings to the dbnet.cfg file on the NetWare DOS partition. After the dbnet.cfg file exists, DBNET uses the settings in it and no longer detects new settings.

The Config option creates a new configuration file using one of the available IP network boards.

The Reset option deletes the dbnet.cfg file and restarts the debug network.

dbstat

Displays debug network statistics. This command also work at the debug console.

syntax: mem | mlid | lsl | arp | ip | udp [conn] | tcp [conn]

mem: Displays memory statistics

mlid: Displays MLID statistics

lss: Displays LSL statistics

apr: Displays APR statistics

ip: Displays IP statistics

udp: Displays UDP statistics

tcp: Displays TCP statistics

conn: The Conn option displays the connection statistics for the protocol specified.

6.2.4 Manually Setting the DBNET Configuration (Optional)

IMPORTANT:Setting the DBNET configuration is completely optional and is not recommended unless it is necessary. It is only necessary when you want to use a different adapter than the one automatically chosen or change any of the default settings.

The dbnet6.nlm automatically configures the network configuration. However, you can use the information in this section to create a dbnet.cfg file manually and store it in the server’s boot directory.

Creating the Dbnet.cfg File

The following show a list of sections and keywords that are accepted in the dbnet.cfg file. The sections and keywords are case sensitive:

Table 6-4 Dbnet.cfg File Sections and Keywords Options and Purposes

Section

Keywords:

Purpose

dbmild

mlid

Debug network module with command line options

dbip

addr

Specify the IP address of debug board

 

mask

Specify the IP network board mask

 

defr

Specify the default IP router

dblst

NOTE:The dblsl section is not required for most network drivers. By default, DBNET will try to determine the packet size required by the network board. For an Ethernet board, DBNET chooses an initial buffer size of 2048 (a maximum size required by some Ethernet drivers); otherwise, the operating system LSL buffer size is used.

min buffers

Specify the minimum LSL buffers to allocate

Default: 128

Range: 128 to 256

 

max buffers

Specify the maximum LSL buffers to allocate

Default: 256

Range: 128 to 256

 

buffer size

Specify the LSL buffer packet size in bytes

Default: operating system LSL buffer size

Range: 632 to 4280

The following is a sample dbnet.cfg configuration file:

[dbmlid] mlid=3C59X FRAME=Ethernet_II SLOT=10005 [dbip] addr=192.65.79.47 mask=255.255.252.0 defr=192.65.79.254