What is a ServerNet Cluster? Revised: 21, May 2004 What's New? Home Frequently Asked Questions How Do I Known Problems With: Manuals Online Help ZPMCONF, ZSCCONF macros Contact Us Help The HP NonStop™ ServerNet Cluster product is an interconnection technology for NonStop S-series servers. This technology enables multiple servers to be connected in a group, or ServerNet cluster. A server or "node" in the cluster can communicate with any other server in the cluster using the ServerNet protocol.
Layered Topology The layered topology is built with the HP NonStop ServerNet Switch (model 6780). The layered topology supports up to 64 nodes and is generally available with the G06.22 RVU. For details about the layered topology, see the ServerNet Cluster 6780 Planning and Installation Guide.
What's New and What's Changed for the ServerNet Cluster Web Site May 21, 2004 A new cause code that can appear in the SCF STATUS SUBNET display has been added to Known Problems With Manuals for the ServerNet Cluster Manual and the ServerNet Cluster 6780 Operations Guide.
The G06.21 RVU supports the HP NonStop ServerNet Switch (model 6780) and the layered topology for ServerNet clusters. For more information, see the ServerNet Cluster 6780 Planning and Installation Guide and the ServerNet Cluster 6780 Operations Guide. December 16, 2002 Known Documentation Problems has been replaced by two separate topics: Known Problems With Manuals and Known Problems With Online Help. Each of these topics is organized by the version of manual or online help system.
❍ The height of the cluster switch enclosure is listed as 37.6 in. (95.5 cm) in the Switch Enclosure Dimensions table in Section 2, "Planning for Installation." The correct height of the switch enclosure is 20.5 in. (52 cm). October 1, 2001 ● Known problems with Documentation. Added the following corrections: ❍ The online help file (SWITCH.CHM) that is included with the TSM 2001C client software is the wrong file. TSM 2001C contains Version 1.
What is a ServerNet Cluster? The HP NonStop™ ServerNet Cluster product is an interconnection technology for NonStop S-series servers. This technology enables multiple servers to be connected in a group, or ServerNet cluster. A server or "node" in the cluster can communicate with any other server in the cluster using the ServerNet protocol. ServerNet clusters extend the ServerNet X and Y fabrics outside the system boundary and allow the ServerNet protocol to be used for intersystem messaging.
Frequently Asked Questions: HP NonStop™ ServerNet Cluster Below is a list of frequently asked questions (FAQ) for the NonStop™ ServerNet Cluster product. The questions are grouped into topics, and the topics are listed in alphabetical order. The table provides links to each question. If you can't find the answer to your question or you want to propose a question for the FAQ, please click the Contact Us link in the main menu to find out how to let us know.
What is an external ServerNet fabric? What are guided procedures? Dial-Outs Documentation Expand How do dial-outs work in a ServerNet cluster? Which manuals document the Nonstop™ ServerNet Cluster product? What does the Expand-over-ServerNet line-handler process do? What is the Expand/ServerNet profile? If two systems are linked by more than one type of Expand line (for example, Expand-over-ServerNet and Expand-over-FOX), which line will be used? Installation Do I have to install new hardware to construc
Are PICs replaceable in the NonStop™ Cluster Switch (model 6770)? Are PICS replaceable in the NonStop™ ServerNet Switch (model 6780)? Power Products What is the function of a UPS? What is the function of an AC transfer switch? What is the function of a power distribution unit (PDU)? SANMAN What does SANMAN do? How often does SANMAN poll each ServerNet II switch? SCF What SCF commands are supported for the ServerNet cluster subsystem (SNETMON)? What SCF commands are supported for the external ServerNet
Cables What kind of cables are required for connections between a node and a cluster switch? Connections between the MSEBs in group 01 of each node and the cluster switches are made using single-mode fiber-optic cables. Connections between cluster switches using the star topologies also use single-mode fiber-optic cables. Connections between cluster switches using the layered topology use either multimode or single-mode fiber-optic cables.
Cluster Size How many nodes are supported in a ServerNet cluster? Up to 24 nodes are supported for the tri-star topology. Up to 64 nodes are supported for the layered topology. Will larger clusters be supported in future releases? Possibly. The ServerNet cluster architecture can theoretically support 1,024 nodes. However, ServerNet clustering relies on Expand, which currently supports no more than 255 nodes. Is it possible to have a one-node cluster? Yes.
or more other computer systems in a network. Nodes in a ServerNet cluster are called "ServerNet nodes." Because ServerNet clusters use Expand, a ServerNet node is also an Expand node. What is a Nonstop™ Cluster Switch (Model 6770)? A Nonstop™ Cluster Switch is an assembly consisting of a ServerNet II Switch, an uninterruptible power supply (UPS), and an AC transfer switch. 6770 switches support the star topologies.
cluster, the external fabrics consist of the fiber-optic cables and cluster switches that provide the infrastructure for routing ServerNet packets between ServerNet nodes. Each Nonstop™ S-series server also has internal ServerNet fabrics that route ServerNet packets either to a ServerNet device within the server or out across the external fabrics to another node. What are guided procedures? Guided procedures are software tools that assist in the configuration or replacement of system components.
What does the Expand-over-ServerNet line-handler process do? The Expand-over-ServerNet line-handler process is a process pair that handles incoming and outgoing Expand messages using the ServerNet protocol. Like the Expand-over-FOX line-handler process, Expand-over-ServerNet manages security-related messages. Other messages usually bypass the line-handler process and are handled directly by the message system. A single Expand line is associated with an Expand-over-ServerNet line-handler process.
Add SPRs to G06.19 and G06.20 to obtain support for the layered topology built with 6780 switches (for clusters up to 64 nodes). For information about these SPRs, refer to the ServerNet Cluster Planning and Installation Guide. Are ServerNet clusters customer-installable? Yes. With special training, customers can perform all tasks for installing a ServerNet cluster.
How are the message system paths monitored? SNETMON maintains a matrix of information about the status of processor-to-processor connections between nodes. This information is made available to clients, such as the TSM package, for use in incident analysis. You can view this information using the SCF STATUS SUBNET command. For a local node, SNETMON stores the state of each internal path within that node.
all—single-mode fiber PICs. The NNA FPGA is installed only in the single-mode fiber PIC of port 6 of the MSEB installed in slots 51 and 52 of group 01. The NNA modifies the ServerNet ID for each ServerNet packet entering or exiting the external ServerNet fabrics. How does the NNA modify ServerNet packets for delivery over the external ServerNet fabrics? All ServerNet packets destined for other nodes in a cluster must pass through port 6 of the MSEB in slot 51 or 52 of the group 01 enclosure.
Are PICs replaceable in MSEBs? Yes. PICs installed in an MSEB are Class-3 CRUs. However, they can be replaced only by personnel who have Compaq-provided training. Are PICs replaceable in the Nonstop™ Cluster Switch (Model 6770)? No. PICs installed in the ServerNet II Switch component of the 6770 switch cannot be replaced at the installation site. If any component of the switch requires servicing, the entire switch must be replaced. Are PICs replaceable in the Nonstop™ServerNet Switch (Model 6780)? Yes.
process pair with the process name $ZZSMN. There is a SANMAN process pair on every node in a functioning ServerNet cluster. How often does SANMAN poll each cluster switch? The polling interval depends on the version of SANMAN that is running on your system and the state of the link. If a problem occurs with the link, later product versions of SANMAN poll the switch more often until the link is repaired. SANMAN Version T0502 (G06.09) T0502AAA (G06.09) T0502AAE (G06.12) T0502AAG (G06.14) T0502AAH (G06.
What’s the difference between ServerNet I and ServerNet II? ServerNet I is the first-generation ServerNet protocol. It features 50 MB/s speed, six-port routers, parallel encoding, and a 64-byte maximum packet size. ServerNet II is the second-generation ServerNet protocol. It features 125 (or 50) MB/s speed, 12-port routers, serial encoding, and a 512-byte maximum packet size. Back to Top of FAQ Servers Can a ServerNet cluster include different types of NonStop™ S-series servers? Yes.
or on top of each other. When packaged in switch enclosures, Nonstop™ Cluster Switches can be installed on the floor or on top of a single system enclosure. (They must not be installed on top of two stacked system enclosures). How does a ServerNet II Switch become configured as an X-fabric or Y-fabric switch? Newly ordered ServerNet II Switches are not configured for either fabric.
● ● ● ● The split-star topology, supported for G06.12 (or equivalent SPRs) or later RVUs, allows up to two cluster switches per fabric. The tri-star topology, supported for G06.14 (or equivalent SPRs) or later RVUs, allows up to three cluster switches per fabric. The layered topology, supported for G06.21 (or equivalent SPRs) or later RVUs, allows up to 12 cluster switches per fabric. For all topologies, each cluster switch supports up to 8 ServerNet nodes.
ServerNet II Switch CRU extended for servicing
NonStop™ Cluster Switch
Modular ServerNet Expansion Board (MSEB)
ServerNet Single-Mode Fiber Plug-In Card (PIC)
NonStop™ Cluster Switch Enclosure
How do I ... How do I ... Add a node to a ServerNet cluster? Solution You add a node to a ServerNet cluster by using a guided procedure. You must perform the procedure on a workstation that has a LAN connection to the node you want to add. The guided procedure that you use depends on whether you are using TSM or OSM. (OSM is required for clusters that use the layered topology.
Replace an AC Transfer Switch? You replace an AC transfer switch, a component of a Nonstop™ Cluster switch, by using a guided procedure. Online help for the guided procedure provides detailed instructions on how to physically replace an AC transfer switch. It is recommended but not required that the AC transfer switch be replaced by a service provider. You can perform the procedure on a workstation that has a LAN connection to any node connected to the Nonstop™ Cluster Switch.
Replace a UPS? The uninterruptible power supply (UPS), a component of a Nonstop™ Cluster Switch, is a field-replaceable unit that can be replaced only by specially-trained service providers. Your service provider must use the NonStop S-Series Service Providers Supplement in conjunction with a guided replacement procedure. Detailed instructions on how to physically replace a UPS are available to service providers in the supplement.
Add a NonStop Cluster Switch (model 6770) to each fabric of a ServerNet cluster to upgrade to the split-star or tri-star topology? Replace a NonStop ServerNet Switch (model 6780)? Replace components in a 6780 switch? Use the Add Switch guided configuration procedure (Start>Programs>Compaq TSM>Guided Configuration Tools>Add Switch). The steps for using the procedure depend on the topology you want to upgrade to. In general, you start with two or three functioning ServerNet clusters.
Known Problems With Manuals Click the Links in the Part Numbers column of the table below to display known documentation problems for specific manuals. Detailed problem descriptions and workarounds are provided. Updated May 21, 2004 Manual Title Himalaya S-Series Service Provider Supplement Himalaya S-Series System Expansion and Reduction Guide ServerNet Cluster Manual Part Numbers With Known Problems Released With 429953-001 G06.12 426721-001 G06.09 427555-001 G06.10 429228-001 G06.
Back to Top of Page Himalaya S-Series System Expansion and Reduction Guide (PN426721-001, PN427555-001, and PN429228-001) Problem: A caution on page 1-2 of the guide incorrectly states that if a system you want to resize is a member of a ServerNet cluster and you want to perform this expansion online, you must first remove the system from the cluster. Correction: There is no restriction for expanding or reducing a system online when a system is a member of a ServerNet cluster.
ports on the ServerNet II Switches is included in Section 5 of the G06.10 revision of the ServerNet Cluster Manual (427410-001). For more information see How Do I ... topic. Problem: The height of the cluster switch enclosure is listed as 37.6 in. (95.5 cm) in the Switch Enclosure Dimensions table in Section 2, "Planning for Installation." The correct height of the switch enclosure is 20.5 in. (52 cm). Correction: The correct height of the cluster switch enclosure appears in the G06.
Aborting and restarting SANMAN causes the nodes to correctly pick up the high node numbers (9 through 16) and configure the X and Y NNAs with these new numbers. Workaround: To avoid this problem, use the G06.13 version of the ServerNet Cluster Manual (520440-001). The procedure has been corrected in the G06.13 manual. If you do not have access to the G06.13 manual, perform the "Steps for Merging Two ServerNet Clusters" in Section 4 of the ServerNet Cluster Manual.
The TSM incident analysis (IA) software generates an alarm if one external fabric has two cluster switches while the other external fabric has only one cluster switch. After a second cluster switch is added to an external fabric, the IA checks the peer fabric to determine if it has two cluster switches. If, after 4 minutes, only one external fabric has two cluster switches, the IA generates a Missing Remote ServerNet Switch alarm.
To access the TIM server, set your Default WWW Master Collection under Edit-->Preferences to: http://tex.zcac.compaq.com/cgi-bin/timcfg.exe or http://tex.tandem.com/cgi-bin/timcfg.exe Problem: The height of the cluster switch enclosure is listed as 37.6 in. (95.5 cm) in the Switch Enclosure Dimensions table in Section 2, "Planning for Installation." The correct height of the switch enclosure is 20.5 in. (52 cm). Correction: The correct height of the cluster switch enclosure appears in the G06.
DELAY 5 START PROCESS $ZZKRN.#ZZSMN Aborting and restarting SANMAN causes the nodes to correctly pick up the high node numbers (9 through 16) and configure the X and Y NNAs with these new numbers. Workaround: To avoid this problem, use the G06.13 version of the ServerNet Cluster Manual (520440-001). The procedure has been corrected in the G06.13 manual. If you do not have access to the G06.13 manual, perform the "Steps for Merging Two ServerNet Clusters" in Section 4 of the ServerNet Cluster Manual.
Problem: The manual contains no information about the Missing Remote ServerNet Switch alarm or time limits for connecting the four-lane links. The TSM incident analysis (IA) software generates an alarm if one external fabric has two cluster switches while the other external fabric has only one cluster switch. After a second cluster switch is added to an external fabric, the IA checks the peer fabric to determine if it has two cluster switches.
ServerNet Cluster Manual (520575-001) Problem: In Section 4, "Upgrading a ServerNet Cluster," the "Steps for Merging Two Star Topologies to Create a Split-Star Topology" include sub-steps for aborting all Expand-over-ServerNet lines and stopping the ServerNet cluster subsystem ($ZZSCL) before running the guided procedure for adding a cluster switch (Add Switch). However, the guided procedure will not run if the $ZZSCL subsystem is stopped.
Back to Top of Page Himalaya Cluster Switch Hardware Installation and Support Guide Problem: The manual does not say what type of AC power receptacle is required for the primary and secondary power sources for the cluster switch. Corrections: ● The AC power cord for North America and Japan (T24242) uses a NEMA 5-15P plug, so the power sources must have NEMA 5-15R receptacles. ● The international power cord (T29490) uses a CEE 7 plug, so the power sources must have CEE 7 receptacles.
Known Problems With Online Help Click the Links in the Versions column of the table below to display known documentation problems for specific online help systems. Detailed problem descriptions and workarounds are provided. Updated December 16, 2002 All guided procedures Versions With Known Problems All Add Switch 1.0 TSM 2001B 2.0 Help System Title Released With All TSM releases Configure ServerNet Node 1.0 TSM TSM TSM TSM Replace Switch 1.0 TSM 2001C Troubleshoot ServerNet Fabric 1.
● SWITCH.CHM (supports the Replace Switch Component procedure) Double-click any of the .CHM files to open the HTML Help application independent of the guided procedures interface. To print one or more topics, click the Print button. Back to Top of Page Add Switch Guided Proedure (Versions 1.0 & 2.0) Problem: The online help contains no information about the Missing Remote ServerNet Switch alarm or time limits for connecting the four-lane links.
updated Help file on a TSM workstation, copy the file to the C:\ZSUPPORT\GUIDED PROCEDURE\HELP directory. You can overwrite the current SWITCHADDITION.CHM Help file. Back to Top of Page Configure ServerNet Node Guided Procedure (Version 1.0) Problem: Clicking the Help button on the CNP0001: Internal Error dialog box shows "Page not found" instead of the online help topic for the error dialog box. Workaround: After you have dismissed the error dialog box, do the following: 1.
Troubleshoot ServerNet Fabric Guided Procedure (Version 2.0) Problem: The Loopback Plugs topic lists an incorrect part number (U11674) for the ECL loopback plug. Correction: The correct part number for ordering the ECL loopback plug is 138894. Problem: The version number of the online help file included in the TSM 2001D client software is incorrect. The help file (FABRICTS.CHM) included with the TSM 2001D client software shows the help version as Version 2.0.
ZPMCONF and ZSCCONF ZSCCONF As of the G06.21 RVU, an updated version of the ZPMCONF macro is available . The new version, called ZSCCONF, does not need to be downloaded. It is installed in the $SYSTEM.ZSUPPORT subvolume. For details on running ZSCCONF, see the ServerNet Cluster 6780 Planning and Installation Guide. ZPMCONF ZPMCONF is an example macro that automates the process of adding MSGMON, SANMAN, and SNETMON to the system configuration database.
== This is the main procssing routine. It gets executed once at the == bottom of the TACL macro. [#def mainBody routine |BODY| #output Determining the total of number of processors...
+ allow all errors [#if [#charfindv scf^output 1 "SNETMON"] |THEN| + abort process $zzkrn.#zzscl #delay 200 + delete process $zzkrn.#zzscl] [#if [#charfindv scf^output 1 "SANMAN"] |THEN| + abort process $zzkrn.#zzsmn #delay 200 + delete process $zzkrn.#zzsmn] [#if [#charfindv scf^output 1 "MSGMON"] |THEN| + abort process $zzkrn.#msgmon #delay 200 + delete process $zzkrn.#msgmon] + exit #output #output Verifying that the entries were successfully deleted...
| 2 3 | #set listProcessors (0,1) | 4 5 | #set listProcessors (2,1,3) | OTHERWISE | #set listProcessors (2,5,6,3,7,4) ] scf /name, inline, out [#myterm]/ + allow all errors + add process $zzkrn.#msgmon, autorestart 10, cpu all, & hometerm $zhome, outfile $zhome, name $zim, & priority 199, program $system.system.msgmon, saveabend on, & startmode system, stopmode sysmsg + add process $zzkrn.#zzsmn, autorestart 10, priority 199, & program $system.system.
Contact Us What do you think? Your views on what we do and how we could make things better are important. Please send us your comments or suggestions. We appreciate your input and look forward to hearing from you. ● Click on this link to e-mail your comments regarding any of the HP NonStop publications . A detailed response to your message will be sent as soon as possible.
Help Welcome to the ServerNet Cluster web site! We want to help you get the most out of our site. 1. Which browser should I be using to get the most from your site ? The web site is optimized for version 4 and above of Internet Explorer. 2. How do I know if I am viewing the most up-to-date material? The most up-to-date material is available on the Compaq extranet and the Compaq intranet (Compaq employees only).
