[an error occurred while processing this directive]

HP OpenVMS Systems Documentation

Content starts here

HP OpenVMS System Management Utilities Reference Manual


Previous Contents Index


Chapter 18
SCA Control Program Utility

18.1 SCACP Description

The SCA Control Program (SCACP) utility is designed to monitor and manage cluster communications. It is derived from the Systems Communications Architecture (SCA), which defines the communications mechanisms that allow nodes in an OpenVMS Cluster system to cooperate.

SCA does the following:

  • Governs the sharing of data between resources at the nodes
  • Binds together System Applications (SYSAPs) that run on different Alpha and VAX computers

Historically, LAN cluster port information has been available only in the System Dump Analyzer (SDA) utility and by using the Availability Manager management tool. The ability to start and stop PEDRIVER on a LAN device was provided by SYS$EXAMPLES:SYS$LAVC_START_BUS.EXE and SYS$LAVC_STOP_BUS.EXE. No way existed to prioritize use of LAN devices or individual channels.

SCACP provides an alternative method of collecting cluster management data and exercising management control over cluster communications. OpenVMS Version 7.3 introduced SCACP's ability to manage SCA use of LAN paths. Beginning with OpenVMS Version 7.3-1, you can use SCACP to manage all OpenVMS Cluster interconnects.

18.1.1 Terminology Related to SCACP

Definitions of terms that are related to SCACP are in the following sections.

18.1.1.1 SCS Ports and Circuits

SCA communications mechanisms between nodes are defined in terms of System Communications Services (SCS) ports and circuits:

  • An SCS port is any device that provides SCA communications services.
  • An SCS circuit is an SCS port layer connection that provides a standardized set of services using a reliable port-to-port communication connection between OpenVMS Cluster nodes.

After a circuit is formed between two ports, communication using the SCS services can be established between SYSAPs in the nodes. In a cluster, each port maintains a circuit with every other remote port.

Circuits provide the SCS layer with the following standardized SCS services:

  • Datagram delivery
  • Message delivery
  • Block data read and write operations

Note

Some differences exist in the use of the terms paths and circuits:
  • The SCA architecture specification and OpenVMS code use the term paths to refer to circuits.
  • The SHOW CLUSTER utility and other OpenVMS utilities use the term circuits to refer to what SCA calls paths.
  • SCACP follows the SHOW CLUSTER precedent and uses the term circuits as well.

18.1.1.2 Virtual Circuits

A virtual circuit (VC) is the interconnect-specific transport layer connection within a circuit that provides reliable port-to-port communication. In other words, VCs are the details of a circuit concerned with interconnect-specific reliable data delivery.

Circuits ensure the following:

  • The delivery of messages without duplication or loss
  • The sequential ordering of messages

The concepts of circuits and virtual circuits are so closely related that usually it has been unnecessary to differentiate between them because SHOW CLUSTER provides no visability into the internal operation of a circuit. SCACP, however, makes the differentiation necessary by providing this internal visability. Beginning with OpenVMS Version 7.3, SCACP can be used to manage and display information about the VCs underlying the circuits between LAN ports. Because SCACP displays different types of information about circuits and VCs, its commands must differentiate between them.

LAN cluster communications create virtual circuits using the NI-SCA Transport Protocol to communicate over LAN hardware, providing datagram services.

18.1.1.3 LAN Channels

A channel is a logical communication path between two LAN devices. Each channel between two nodes is determined by a local-remote pair of devices and the connecting network. For example, two nodes, each having two LAN devices, can establish up to four channels. The messages that a particular virtual circuit carries can be sent over any of the channels connecting the two nodes.

The LAN cluster driver, PEDRIVER, builds reliable virtual circuits using channels that the LAN adapters (devices) and the network connections define. It then uses these VCs to provide circuits to SCS.

18.1.1.4 Channels and Virtual Circuits

The differences between channels and virtual circuits are the following:

  • Channels are LAN paths providing datagram service.
  • NI-SCA port-to-port virtual circuits are layered on channels and provide error-free paths between nodes.

Multiple channels can exist between nodes in an OpenVMS Cluster system, but only one LAN-based virtual circuit can exist between any two nodes at a time.

18.1.2 New Cluster SCA Circuit and Port Functionality

The following sections explain new functionality available for SCA circuits and ports and new support for SCS dynamic load class. In addition, you can independently enable or disable checksumming on individual PEdriver virtual circuits (VCs).

18.1.2.1 Ability to Set Port and Circuit Priorities

Beginning with OpenVMS Version 7.3-1, you can exercise management control over the circuits chosen to be used for SCS connections. This control allows you to override the automatic selection of the circuit with the highest load class value.

To override automatic circuit selection, assign a management priority value to a specific circuit or SCA port. (A circuit's current priority value is the sum of the local port's management-assigned priority and the management priority value assigned to that circuit.)

Connections are assigned to a circuit with the highest priority. If multiple circuits have the highest priority value, then the circuit with the highest load class is selected.

A change in a circuit's current priority has one of the following effects:

  • If a circuit's new current priority value is higher than another circuit's current priority, the connection is moved to the circuit with the higher current priority.
  • Similarly, if a circuit's new current priority value is lower than another circuit's current priority, the connection moves to the circuit with the highest current priority.

CIRCUIT MANAGEMENT PRIORITY IS VOLATILE

Whenever a circuit is closed, its management priority setting is lost. This is because the data structure containing information about a circuit is deallocated each time a circuit is closed. When a circuit is reopened, the structure is initialized with default values. Thus, circuit management priority does not propagate across VC closures.

18.1.2.2 Ability to Enable/Disable PEdriver Checksumming

You can use SCACP to enable or disable checksumming on a per-VC basis. For example, in a disaster tolerent cluster, you might want to enable only checksumming on VCs to nodes at the remote site to ensure that failure of a LAN device's checksumming function resulting in corrupted packets would not propagate to the remote site.

18.1.2.3 SCS Dynamic Load Class Support

Prior to OpenVMS Version 7.3-1, the load class of SCS circuits was determined only by the port's hard-coded load class value. As a result, CI or DSSI circuits were chosen over a GigaBit Ethernet circuit. Beginning with OpenVMS Version 7.3-1, PEdriver dynamically updates the load class value of its SCS circuits to reflect the performance of the underlying LAN path or paths currently in use.

If the circuits have the same priority, a change in a circuit's load class has one of the following effects:

  • If a circuit's new load class value is higher than another circuit's current load class, the connection is moved to the higher load class circuit.
  • Similarly, if a circuit's new load class value is lower than another circuit's load class, the connection moves to the circuit with the highest load class.

18.1.3 Managing Cluster Ports and Circuits

Beginning with V7.3-1, SCACP provides the ability to display information about one cluster interconnect's local ports and their circuits with remote ports.

SCACP port and circuit data is intended to provide the information necessary to exercise management control over ports and circuits. SCACP is not intended to replace the wealth of data that SHOW CLUSTER provides about ports and circuits. The SHOW CLUSTER and SCACP utilities are intended to be used together to manage cluster communications.

SCACP port and circuit data show the following:

  • SCACP port data shows an overview of a particular port's characteristics.
  • SCACP circuit data shows the characteristics and the status of SCS communications with other nodes in the cluster.

You can also manage cluster communications by assigning a priority value to individual ports or circuits. See the SET CIRCUIT and SET PORT command descriptions.

18.1.4 Managing LAN Cluster Ports

To manage LAN cluster ports, you can use common port and circuit commands. Additional commands exist for LAN port VCs, channels, and LAN devices on nodes in the cluster:

  • VC data shows detailed internal information about the characteristcs and operation of the NI-SCA transport layer connection underlying the circuit between the local and remote PEDRIVER ports.
  • Channel data shows the characteristics of each LAN communications path and shows how suitable each channel is for use by the virtual circuit.
  • LAN device data shows low-level local LAN device characteristics, counters, and errors.

SCACP allows you to set channel and LAN device priority. SCACP also allows you to start and stop PEDRIVER on LAN devices.

Using the PEDRIVER Event-Tracing Facility

The LAN cluster port driver, PEDRIVER, includes an event-tracing facility that collects a small amount of information for each defined event and saves it in a buffer associated with the virtual circuit or channel. (Any event not associated with a particular virtual circuit or channel is saved in a global PEDRIVER trace buffer.)

The event trace data is used when debugging, performing dump analysis, and looking at detailed aspects of PEDRIVER operation.

Note

The TRACE commands are reserved for use by OpenVMS Engineering and HP Services under OpenVMS Engineering direction only. Trace commands and output are subject to change from release to release.

18.1.5 Troubleshooting Cluster Communications

You can use SCACP to diagnose cluster communications problems. The appendix "Troubleshooting the NISCA Protocol" in HP OpenVMS Cluster Systems provides troubleshooting strategies to help cluster or network managers use SCACP and other tools to pinpoint network-related cluster communications problems.

18.1.6 Related Documents

For additional information on the topics covered in this chapter, refer to the following documents:

  • HP OpenVMS Cluster Systems
  • Guidelines for OpenVMS Cluster Configurations

18.2 SCACP Commands

SCACP commands are provided for the following types of functions:

  • Display
  • Port selection
  • Circuit selection
  • Channel operation and selection
  • LAN device operation and selection
  • Trace
  • Miscellaneous: Help, Spawn, Execute, and Exit

The SCACP commands are shown in Table 18-1.

`

Table 18-1 SCACP Commands
Command Function
SCACP Display Commands Qualifiers
SHOW CHANNEL Displays PEDRIVER channel information.
SHOW CIRCUIT (nodename) Shows information about all circuits between this node and other cluster nodes, including LAN circuit information.
SHOW LAN_DEVICE Displays PEDRIVER LAN device information.
SHOW PORT Displays information about all SCA ports on the node, including the LAN port, PEA0.
SHOW VC Displays PEDRIVER virtual circuit information.
SCACP Operation Commands
SET CHANNEL Allows a user to set PEDRIVER channel management options.
SET CIRCUIT Allows a user to set a management priority value for the selected circuit or circuits.
SET PORT Allows a user to set a management priority value for the selected port or ports.
SET VC Allows a user to set PEDRIVER virtual channel options.
SCACP LAN Device Operation Commands
SET LAN_DEVICE Sets PEDRIVER LAN device management options.
START LAN_DEVICE Starts PEDRIVER on the specified LAN devices.
STOP LAN_DEVICE Stops PEDRIVER on the specified LAN devices.
SCACP Trace Commands
These commands are reserved for HP use only.
SET TRACE Sets PEDRIVER event tracing options.
SHOW TRACE Displays PEDRIVER event tracing information.
START TRACE Displays PEDRIVER event tracing.
STOP TRACE Stops PEDRIVER event tracing.
SCACP Miscellaneous Commands
HELP Displays help data.
SPAWN [command] Spawns and executes a DCL command.
@filename Executes command file.
EXIT Exits SCACP.

18.3 SCACP Usage Summary

The SCA Control Program (SCACP) is a cluster management utility that performs certain privileged cluster communications management functions.


Format

RUN SYS$SYSTEM:SCACP


Parameters

None.

Description

To invoke SCACP, enter the following command at the DCL prompt:


$ RUN SYS$SYSTEM:SCACP

SCACP displays the following prompt, at which you can enter SCACP commands using the standard rules of DCL syntax:


SCACP>

To exit SCACP and return to the DCL command level, enter the EXIT command at the SCACP> prompt or press Ctrl/Z.

Note

OpenVMS Version 7.3 and later require SYSPRV privilege to issue SCACP commands.

SCACP also requires you to have the following privileges:

  • A minimum of DISPLAY privilege is required to issue commands that display information or influence SCACP execution (that is, SHOW, HELP, SPAWN, EXIT, and so on.)
  • SYSPRV privilege is reqired to issue commands that change cluster communications operations (that is, SET, START,and STOP).

Example


$ CREATE COUNT.COM
SHOW LAN_DEVICE/COUNTERS
SPAWN WAIT 00:01:00
@COUNT

[Ctrl/Z]
$ RUN SYS$SYSTEM:SCACP
SCACP> @COUNT
      

This example creates and runs a command procedure, COUNT.COM, which displays device counters once a minute.

EXIT

Stops execution of SCACP and returns control to the DCL command level. You can also enter Ctrl/Z at any time to exit.

Format

EXIT


Parameters

None.

Qualifiers

None.

Example


SCACP> EXIT
$
      

This command stops execution of SCACP and returns control to the DCL command level.

HELP

Provides online help information about the SCACP utility.

Format

HELP [topic]


Parameter

topic

Specifies a subject for which you want information: SCACP command or command plus command keyword. If you enter the HELP command with a command name only, such as HELP SET, SCACP displays a list of all of the command keywords used with the SET command.

Qualifiers

None.

Example


SCACP> HELP SET TRACE
      

The HELP command in this example displays information about the SET TRACE command.

SET CHANNEL

Sets CHANNEL management parameters, currently limited to priority values.

Note

If a LAN device is disabled for use by cluster communications, all channels associated with that device are deleted. This means that all management settings for that device and its associated channels will be deleted.

Format

SET CHANNEL nodename


Parameter

nodename[,...]

Includes channels to specific nodes, which you can use wildcards to specify. Each node name can be accompanied by optional qualifiers to specify local and remote device names. If no local or remote device name is specified, all channels to the specified node name are included.

Use the SHOW CHANNEL command to display node names and local and remote device names.


Qualifiers

/EXCLUDE=(nodename[,...])

Excludes channels to specific nodes, which you can use wildcards to specify. Each node name can be accompanied by optional qualifiers to specify local and remote device names. If no local or remote device name is specified, all channels associated with the specified node are included.

/LOCAL_DEVICE=(landevicename[,...])

Specifies a LAN device that identifies the local end of the channel; you can use wildcards to specify LAN devices.

Use the SHOW CHANNEL command to display node names and local and remote device names, or use the SHOW LAN_DEVICE command to display device names.

/PRIORITY=n

Sets the management priority value for channels to selected nodes. n can be a value between -128 and +127. Suggested values are:
2 to cause channels to be preferred
-2 to exclude channels

Note

Starting with OpenVMS Version 7.3-2, a channel whose priority is -128 is not used for cluster communications. The priority of a channel is the sum of the management priority assigned to the local LAN device and the channel itself. Therefore, you can assign any combination of channel and LAN device management priority values to achieve a total of -128.

Caution

If you set the priority of all channels to -128, you will totally disable use of the LAN for cluster communication. This can cause the system to CLUEXIT.

/REMOTE_DEVICE=(landevicename[,...])

Specifies a LAN device that identifies the remote end of the channel; you can use wildcards to specify LAN devices.

Use the SHOW CHANNEL command to display node names and local and remote device names, or use the SHOW LAN_DEVICE command to display device names.

Use the SHOW LAN_DEVICE command on the remote node to display remote device names.


Examples

#1

SCACP> SET CHANNEL/PRIORITY=3 NODE5
      

The command in this example sets the priorities of all channels to node NODE5 to 3.

#2

SCACP> SET CHANNEL/LOCAL=EWA/REMOTE=EWB -
_SCACP> NODE10,NODE15/L=F*/R=F*,NODE20/PRIORITY=10
      

The command in this example is equivalent to the following command:


SET CHANNEL NODE10/L=EWA/R=EWB,NODE15/L=F*/R=F*,NODE20/L=*/R:*/PRIORITY=10

This command sets the priority of the following channels to 10:

  • To node NODE10, the channel with local device EWA and remote device EWB
  • To node NODE15, the channels with local LAN devices starting with F and remote LAN devices starting with F
  • All channels to node NODE20


Previous Next Contents Index