DIGITAL Network Process Failover (NPF) Version 1.0 RELEASE NOTES September 9th 1998 Part Number: AA-RF2DA-TE About This Document This document describes important additional information about installing, configuring and using DIGITAL Network Process Failover V1.0 You are strongly advised to read this document carefully before proceeding to read the Installation Guide. Hardware requirements This section outlines special hardware considerations for running DIGITAL Network Process Failover V1.0 Supported Processors As described in the Software Product Description (SPD) 70.43, the following processor types have been certified for use with DIGITAL NPF 1.0: - AlphaServer 800 - Embedded AlphaServer 800 Subrack (Custom Systems Packaging) - AlphaServer 4100 SMP Memory Size Requirement As stated in the SPD 70.43, the minimum required memory size is 128MB per system. Minimum Firmware revision Minimum Firmware revision required to run Digital Network Process Failover is 5.2. Please contact your DIGITAL representative for details about obtaining the latest firmware revision. Configuration requirements Number of systems - The minimum number of interconnected systems running Digital Network Process Failover is 2, both acting as pilot systems. - The maximum number of interconnected systems running Digital Network Process Failover is 4. If Digital Network Process Failover Manager (npfmgr) is used, two of these systems must be configured as pilot systems. These pilot systems must be configured to operate in a TruCluster Available Server environment. See the Digital Network Process Failover Installation Guide and the TruCluster Available Server Installation Guide. Network installation When using Digital Network Process Failover Manager (npfmgr), TWO network links must be installed between the pilot systems, to form a primary network and a backup network. Satellite/Non- pilot systems should be interconnected through the backup network. The Primary network is used exclusively by TruCluster Available Server Environment although the backup network is shared within the Digital Network Process Failover platform. Shared Disk Configuration The two pilot systems of the platform must share disk devices using shared SCSI bus technology. The shared SCSI devices installation must comply with TruCluster ASE hardware requirements. A minimum of 2 disks/2 SCSI buses is required to achieve storage redundancy. Digital recommends shared disk capacity to be at least 1GB. Console Variables Setup In order to Digital Network Process Failover to operate, it is mandatory that the following console variables be set to the following values: AUTO_ACTION RESTART For more information, refer to the System's User Guide for your Processor. Software Requirements This section outlines the Software considerations for DIGITAL Network Process Failover Release 1.0 Minimum Operating System version Digital UNIX Revision 4.0D (Rev. 878) is required for Digital Network Process Failover components to operate. Minimum Patch Release Please consult your DIGITAL Representation for information on how to obtain the following patches. - The minimum patch release required for Digital Network Process Failover to operate is BL10. - You must also obtain and install the following patch: - ALC-07038_V40D This patch contains the following fixes: - Atomic Data Logging (ADL) Performance - NFS Local Access Performance (via ADL) - Panic due to: pmap_change_wiring - If you are using TruCluster 1.5, you must obtain and install TCR1.5 Patch 12.00 Sysconfig parameters Sysconfig parameters are automatically modified during the installation procedure. Refer to the Digital Network Process Failover Installation guide for more information. File System Digital Network Process Failover components and user processes EXCLUSIVELY use the AdvFS file system to allow for data consistency. Therefore, appropriate AdvFS configuration must be performed on all disk devices of each system. Refer to the Digital UNIX Administration Guide and Digital Network Process Failover Installation Guide for more information. If AdvFS file access is to be performed over NFS, appropriate NFS configuration should be performed including proplistd setup and proplist mount option. Refer to the Digital UNIX Network Administration Guide for more information. System Files /etc/hosts must be the same on all systems of the platform. It must contain the node names and address of all the systems that are members of the platform. Do not use NIS or BIND for address resolution for systems that are part of the platform. These addresses MUST be resolved locally. Log and Trace Files - The log files are located in /var/adm/syslog.dated. In order to retrieve messages regarding Digital Network Process Failover components, you should extract messages containing NEC_CM, NPF_, npf_, NCS_ or cs_tcp. - When enabled, trace files are located in /var/tmp. Default mode is trace enabled. Refer to the Digital Network Process Failover Manager Administrator Guide for information on how to change trace mode. Documentation Instead of postscript files, all documents are now supplied in Adobe Acrobat (.pdf) format. To use Adobe Acrobat Reader on DIGITAL UNIX V4.0D, you must first install it from the Layered Products Documentation CD-ROM by executing the script setup.scp in the setup directory. Programming Tips · All SCIMs must be compiled and linked with pthreads and real-time options. · All libraries used by SCIMs must be thread safe. · Modules calling NPF services cannot be packaged in libraries (.a or .so) but should be built as an object module (.o) linked with the SCIM image. · Standard Output (stdout) and Standard Error (stderr) are redirected to /dev/console by default or to /dev/tty if npf_init is started with the -i option. The default output must be modified by the system administrator to the appropriate device or file in /opt/NPF100/sbin/init.d/npf script file. · To enable SCIM debugging the adm_lock flag must be set to the appropriate value. Refer to Section 2.4.6 of the NPF Manager Administrator Guide. Known Problems and Restrictions This section outlines known issues regarding this release of DIGITAL Network Process Failover. Double-fault At platform level, and at process level, simultaneous double-faults will cause unpredictable results. Double-fault cases are NOT supported by this product. Loading/Unloading Kernel Components Too many manual loads/unloads of the Network Process Failover base kernel components on the same station may lead to an ENOMEM status during components installation. Periodic system reboot is recommended to avoid this problem. XMA - If there is no more room in memory to load and start a SCIM, XMA will terminate with an exit status of 0 - If there are too many messages indicating traffic congestion, some messages may be lost. DM - ENOMEM will occur if all of the configured system messages buffers are consumed. - Deallocating a record will send a trigger message with incorrect contents NPF Manager - Scenario Mode is not supported - The Reconfiguration Mode NXA_AUTORECONF_MOD = 0 is not supported when using NPF Manager. - Automatic calculation of the id of an IFAP is not implemented in this Release. This applies to create fap and create sci.