| Netra™ High Availability Suite 3.0 1/08 Foundation Services Glossary | 
   |
The priority of the Foundation Services processes compared to other system processes. The actual priority of a process is a function of its base priority and its relative priority:
actual priority = base priority + relative prioritySee also, base priority and relative priority.
An addressing schema that assigns each peer node an IP address for the following interfaces:
An error scenario in which a cluster restarts with a newly elected master node that is assumed to have highly available data that is synchronized with the previous master node, when, in fact, it is not. When the previous master node comes back (as new vice-master node), the latest updated data on the node will be lost.
Refers to the probability that a service is available for use at any given time. See also, redundancy, reliability, and serviceability.
Also in this context, one of the proposed data management policies. See also, data management policy.
One part of the actual priority of the Foundation Services processes:
A partition that contains the scoreboard bitmap for a replicated partition on the same disk. There is one bitmap partition per replicated partition. In SNDR, bitmap partitions are referred to as “bitmap volumes.” For a listing of all other partition types, see also, partition.
On the Linux OS, channel bonding refers to a kernel driver that aggregates multiple network links into a single link. The bonding driver sends data presented to the aggregated single link through one or more network links, depending on the bonding driver configuration and which link is up.
On the Solaris OS, Solaris trunking is similar to channel bonding for selected dual/quad Ethernet cards.
An interface to multiple network links aggregated as a single link and managed by the bonding driver.
The method of starting a diskless node at cluster startup. See also, DHCP client ID, DHCP dynamic, and DHCP static.
CGTP is a mechanism added at the IP level that makes IP networks reliable. A data packet sent to a CGTP address (virtual) will be duplicated and then transmitted through two IP addresses (not virtual). Every IP address corresponds to a network interface connected to a network link. The two network links must be two completely distinct routes (no common switch or router can be used on these links).If data packets are lost on one link, those sent on the second link should reach the destination address without any latency. Powerful filtering mechanisms are added to the receiver site to withdraw the duplicate packets with a limited impact to performance.
The transfer of statistics from the Node Management Agent (NMA) on each of the peer nodes into the namespace of the master node. Cascading provides the NMA of the master node with a view of the entire cluster.
A virtual physical interface over which IP packets are transferred using the CGTP. See also, Carrier Grade Transport Protocol and virtual physical interface.
A group of peer nodes that are connected by a network. To run the Netra HA Suite Foundation Services, a cluster must have the following characteristics:
A server (master-eligible) node elected as the vice-master node
All other peer nodes are client (master-ineligible) nodes, which are either dataless or diskless nodes
Data replication over IP or shared disks between the master node and vice-master node
See also, client node, dataless node, diskless node, master node, master-eligible node, master-ineligible node, peer node, and vice-master node.
The number that identifies a cluster. Also called domainid. The clusterid is the same for each peer node in a cluster. See also nodeid.
Cluster Membership Manager is a service that manages the cluster membership and mastership. The CMM performs the following tasks:
The private network over which peer nodes communicate with each other. See also external network.
A configuration file that stores the membership and configuration information for all peer nodes in a cluster. A node must have an entry in the cluster_nodes_table file to be part of a cluster. There is a cluster_nodes_table on both master-eligible nodes in the cluster.
A peer node that boots from its local disk. A dataless node runs customer applications locally, but stores and accesses reliable (highly available) data through the cluster network on a replicated or shared partition of the master node, using the Reliable File Service of the Foundation Services. A dataless node cannot be the master node or the vice-master node. For a listing of all other node types, see also, node.
Keeping data correct, complete, whole, and consistent with the intent of the data’s creators. It is achieved by preventing accidental or unauthorized insertion, modification or destruction of data in a database.
The data management policy of a cluster determines how the cluster behaves when a failed vice-master reboots in a cluster that has no master node. The policy you choose depends on the availability and data-integrity requirements of your cluster. The three data management policies are integrity, availability, and adaptability. The integrity policy is chosen when data integrity must be privileged; availability is chosen when data availability must be privileged (however, note that some data can be lost); adaptability is chosen when you are relying on the state of synchronization between the two master nodes to choose the behavior of the vice-master node.
A partition that contains data. If data contained in the partition must be highly available, the data partition must be either a replicated partition (if the cluster is configured to use data replication over IP) or a shared partition (if the cluster is configured to use shared disks).
See also IP replication and shared disk. For a listing of all other partition types, see also, partition.
Optional hardware used to develop applications using the Netra HA Suite APIs. This hardware can run the Solaris OS or Linux distributions. Refer to the Netra High Availability Suite 3.0 1/08 Foundation Services Getting Started Guide for more information about using a development server.
A boot policy that associates a CLIENT_ID string with a diskless node. See also, DHCP dynamic and DHCP static.
A boot policy that creates a dynamic map between the Ethernet address of a diskless node and an IP address taken from a pool of available addresses. See also, DHCP client ID and DHCP static.
DHCP dynamic is not recommended to be used with Reliable Boot Service, as it is no longer supported for this type of Reliable Service.
A boot policy that maps the Ethernet address of a diskless node to a fixed IP address. See also, DHCP client ID and DHCP dynamic.
A link between the serial ports on the master-eligible nodes. The direct link helps prevent the occurrence of split brain when the network between the master node and vice-master node fails.
A peer node that contains at least one disk on which applications can run and information can be permanently stored. Master-eligible nodes are diskfull nodes. Dataless nodes are not considered to be diskfull nodes. For a listing of all other node types, see also, node.
A peer node that does not have a local disk or is not configured to use its local disk. Diskless nodes boot through the network, using the master node as a boot server. Diskless nodes run customer applications locally, but store and access reliable (highly available) data through the cluster network on a disk partition of the master node, using the Reliable File Service of the Foundation Services. A diskless node cannot be the master node or the vice-master node. For a listing of all other node types, see also, node.
A master-eligible node that cannot participate in an election for the master role or vice-master role. Disqualified is a state used by the Cluster Membership Manager.
DRBD is open source software for data replication over IP. This software is used on Linux by the Reliable File Service when data is shared between master nodes using IP replication. On the Solaris OS, SNDR is similar to DRBD. See also, IP replication, Reliable File Service, and Sun StorEdge™ Network Data Replicator.
Services that run on all peer nodes. The distributed services include the Cluster Membership Manager, the Node Management Agent, and the Process Monitor Daemon. See also, highly available services.
(domain identification number) The number that identifies a cluster. The domainid is the same for each peer node in a cluster. Also called clusterid. See also, nodeid.
The simultaneous failure of both master-eligible nodes, or the simultaneous failure of both redundant networks.
The percentage of time that a system is unavailable, including attributable, scheduled, unscheduled, total, and partial system outages. Downtime also includes outages caused by operational error. See also, availability and outage.
DHCP is a TCP/IP protocol that enables machines to obtain IP addresses statically or dynamically from centrally administered servers.
A 48-bit address used to direct datalink layer transactions. The Ethernet address is synonymous with the MAC address.
An IP address visible on an external network, that is assigned to a logical interface or physical interface on a peer node. See also, floating external address.
A network that is not the intra-cluster network and from which cluster nodes can be reached. An external network can be Ethernet, ATM, or any other supported network type. See also, cluster network.
EAM is a service that configures floating external addresses for external interfaces on the master node. After a failover or switchover, the EAM configures floating external addresses for external interfaces on the new master node. EAM also monitors the state of either the IPMP groups on the Solaris OS or the bonding interfaces on the Linux OS, and triggers a failover when an IPMP group or a bonding interface fails. See also, bonding, bonding interface, floating external address, Internet Protocol Multipathing, master-eligible node, failover, and switchover.
The transfer of services from the master node to the vice-master node when the master node fails. The vice-master node must have all of the necessary state information to take over at the moment of failover. The vice-master node expects no cooperation or coordination from the failed master node. See also, switchover.
A logical address triplet for the node holding the master role. Diskless nodes and dataless nodes access services and data on the master node through the floating address triplet. See also address triplet.
A logical IP address visible on an external network that is assigned to an interface on the node holding the master role.
The basic set of services provided by the Netra HA Suite software, which includes the following Foundation Services:
Applications that are aware of the availability of nodes and resources. The resources can be other application components, communication end points, processes, or groups of resources. An application that is HA-aware is designed to recover from a failure.
Applications that are unaware of the availability of nodes and resources. To enable these applications to recover from a failure, a dedicated agent (interfacing with CMM) must be provided to manage application recovery in the event a failure occurs. NSM can be used to write such an agent.
An IP packet that is periodically multicast over the cluster network through each of the two physical interfaces of each peer node. When a heartbeat is detected through a physical interface, it indicates that the node is reachable and that the physical interface is alive. A probe mechanism is another way to design this mechanism.
Services that run on the master node only. The Reliable Boot Service and Reliable File Service are highly available services. If the master node or one of these services on the master node fails, a failover occurs. See also, distributed services.
The ability to add client (master-ineligible) nodes to a cluster to increase the processing capacity of the system.
The second part of an Internet address. The host part of an IP address identifies a node on a given network. See also netmask and network part.
The removal and replacement of a hardware component or board without shutting down the entire system.
A peer node that is using the Foundation Services and can communicate with other peer nodes. An “in node” can be a master-eligible node, a diskless node, or a dataless node. The master node and vice-master node must always be “in nodes.” See also, out node. For a listing of all other node types, see also, node.
The hardware required to install the OS, the Netra HA Suite software, and some known patches on a Foundation Services cluster. The installation server can be any system capable of running the Solaris or Linux Operating Systems. Refer to the Netra High Availability Suite 3.0 1/08 Foundation Services Getting Started Guide for information about which OS combinations are supported between the installation server and the cluster.
In this context, one of the proposed data management policies. See also, data management policy.
On the Solaris OS, IPMP is a service that allows the failover of several addresses from a failed interface to a working one. Addresses move between interfaces belonging to the same IPMP group. The floating external address can be managed by IPMP. See also, bonding, bonding interface, External Address Manager, and floating external address.
The copying of data from the master node to the vice-master node through the cluster network (sending data over the IP protocol). Through IP replication, the vice-master node keeps an up-to-date copy of the data on the master node. IP replication is also known as shared nothing mechanism, as it is a way to keep master and vice-master synchronized without sharing any hardware (disks). See also, replicated partition and synchronization.
JDMK is the fundamental framework for developing
distributed Java
management
applications.
JMX are tools for building distributed, Web-based, modular and dynamic solutions for managing and monitoring devices, applications, and service-driven networks.
A partition located on a disk that is private to the node (also known as a local disk), such as a boot partition located on the system disk. For a listing of all other partition types, see also partition.
A supplementary address on a physical interface or virtual interface of a node. One physical interface or virtual interface can have many logical addresses. See also, physical address and virtual address.
discrete logical grouping with its own operating system, resources, and identity within a single computer system. Also referred to as LDom.
provides a CLI to create and manage logical domains and allocate resources to domains
A logical interface configured on a physical interface or virtual interface of a node. A logical interface on an hme0 or cgtp0 interface might be called hme0:1 or cgtp0:1.
LVM is an open source software for disk volume management on Linux. This is a software module for Reliable File Service when using shared disk(s). On the Solaris OS, SVM is similar to LVM.
A peer node (master eligible) that coordinates all cluster membership. The primary instance of External Address Manager, Reliable File Service, Reliable Boot Service, and Node State Manager run on the master node. If a floating external address is to be used, it must be defined on this node. There is only one master node in a cluster. For a listing of all other node types, see also, node.
A peer node that can be elected as the master node or the vice-master node. Master-eligible nodes are always diskfull. For a listing of all other node types, see also, node.
A peer node that cannot perform the role of master node or vice-master node. Also called the client node or satellite node. Master-ineligible nodes are always diskless or dataless. For a listing of all other node types, see also, node.
(management bean) A Java interface that conforms to design patterns that expose attributes and operations to the Node Management Agent (NMA). These attributes and operations enable the NMA to recognize and manage the MBean.
MBeans are defined by the JMX specification.
For more information, go to http://java.sun.com.
A volume manager virtual device, referred to as a logical volume on Linux.
There are several types of metadevices. Netra HA Suite software makes use of two of them:
A partition mirror of a local, replicated, or shared partition. On the Solaris OS, you can only define a mirror partition using SVM. On Linux, you can only define a mirror partition using Linux software RAID. For a listing of all other partition types, see also, partition.
MV CGE Linux is one of three versions of the Linux distribution provided by MontaVista. See http://www.mvista.com for more information.
Corresponds to a daemon or a group of daemons launched by a startup script. See also, probe mechanism and startup script.
A number that defines how many of the upper bits of an IP address identify a network. In class C addresses, the netmask is 24 bits. The upper 24 bits of a class C address identify the network, and the lower 8 bits identify the node. See also, network part and host part.
The upper part of an IP address. The network part of an IP address identifies the network on which a node resides. See also, host part and netmask.
A tool that facilitates the installation and configuration of the Netra HA Suite software and, optionally, the OS and patches on a cluster. See also, installation server.
The basic unit of hardware in a cluster on which the Foundation Services run. A node is either a blade from a blade server or a rack-mounted server. A node can also contain some secondary storage in the form of flash or compact flash memory.
See also the following types of nodes: client node, dataless node, diskfull node, diskless node, in node, master node, master-eligible node, master-ineligible node, out node, peer node, nonpeer node, predefined node, satellite node, server node, and vice-master node.
(node identification number) The number that identifies a node within a cluster. See also, clusterid and domainid.
NMA is a service that monitors cluster statistics, initiates a switchover, changes the recovery response for daemon failure, and listens for notifications of some cluster events. The NMA is a JMX compliant management agent based on the Java Dynamic Management Kit.
NSM is a service that executes user-provided scripts when it becomes master or vicemaster. See also, external address, master node, failover, and switchover.
A node that is not configured as a member of a cluster. A nonpeer node can communicate with one or more peer nodes to access resources or services provided by a cluster. See also, peer node. For a listing of all other node types, see also, node.
Packages that are not shared between the master node and the vice-master node. Nonshared packages must be added to both master-eligible nodes. See also, shared packages.
An information message sent by the nhcmmd daemon on a local node to services or applications registered to receive it.
An event that impairs the ability of a system to operate at its rated capacity for more than 30 seconds.
A peer node on which the Foundation Services are not running and that cannot be reached by the master node. See also in node. For a listing of all other node types, see also, node.
A physical division of a disk or a virtual division of one or more disks. For virtual divisions of disks, partitions are always referred to as soft partitions, which are used through a volume manager. See also, bitmap partition, data partition, local partition, mirror partition, replicated partition, shared partition, and soft partition.
A node in a cluster that is configured to run the Foundation Services. See also nonpeer node. For a listing of all other node types, see also, node.
A network interface card (NIC). The name of the interface depends on the type of network card that is used. A node could have network interfaces called hme0 and hme1.
A node that is defined as part of a cluster but that has not been physically connected to the cluster. A predefined node is a peer node with an out-of-cluster role. For a listing of all other node types, see also, node.
PMD is a service that monitors daemons launched by startup scripts. These daemons can be from Netra HA Suite, operating systems, or companion products. See also, startup script.
The restoration of system operation to full capacity or partial capacity. During recovery, a system performs the following tasks:
RPM Package Manager is a package management system primarily intended for Linux. RPM refers both to a package format, and to a package installation tool.
The provision of a backup node to take over in the event of failure. The Foundation Services use the 2N redundancy model. See also, availability, reliability, and serviceability.
The priority of one of the Foundation Services processes compared to another. In descending order of priority, Netra HA Suite daemons have the following relative priority: nhpmd > nhwdtd > nhprobed > nhcmmd > nhcrfsd > nheamd > nhnsmd. See also, actual priority and base priority.
The measure of continuous system uptime. See also availability, redundancy, and serviceability.
RBS is a service that uses the Dynamic Host Configuration Protocol (DHCP) and other Netra HA Suite services to ensure the boot of diskless nodes regardless of software failures or hardware failures.
RFS is a service that provides a mounted file system to make data on the master node accessible to other cluster nodes. Reliable File Service replicates (with IP replication) or mirrors (with shared disks) disk-based data on the master node to the vice-master node and reconfigures the floating address triplet after failover or switchover. See also, Distributed Replicated Block Device, IP replication, shared disk, and Sun StorEdge™ Network Data Replicator.
A partition located on local disk(s) of a master-eligible node (MEN) and replicated on local disk(s) of the other MEN. The remote copy is kept synchronized by replicating data through IP over the cluster network. See also, IP replication. For a listing of all other partition types, see also, partition.
A membership role allocated by the Cluster Membership Manager. See also, in node, out node, master node, vice-master node.
Feature that allows an upgrade from one version of the Netra HA Suite software to a newer version, one cluster at a time, without needing to take any cluster offline. Note: This feature is not supported in versions previous to 2.1 6/03, and cannot be used to upgrade the operating system.
A parameter that represents the priority with which competing services run. The scheduling parameter is defined in the nhfs.conf file. See also actual priority, base priority, and relative priority.
Bitmap in memory or bitmap in a partition (bitmap partition) that identifies which data blocks on a replicated partition are modified while the vice-master node is out of service. See also IP replication and bitmap partition. Only applicable on the Solaris OS when using the SNDR tool. See also, Sun StorEdge™ Network Data Replicator.
The SA Forum is a consortium of communications and computing companies working together to develop and publish high availability and management software interface specifications. For more information, see http://www.saforum.org
logical domain that provides devices such as virtual switches, virtual console connectors, and virtual disk servers to other logical domains.
The probability that a service can be restored within a specified period of time following a service failure. See also availability, redundancy, and reliability.
Disk or disks in a multi-hosted disk bay separate from the master and vice-master nodes, containing data shared by the cluster. This is an alternative to IP replication for sharing data between master and vice-master nodes. See also, IP replication and Reliable File Service.
Packages that are located on the master node and replicated on the vice-master node, or located on the shared disk. Shared packages can include the Netra HA Suite packages and user applications. See also, nonshared packages.
A soft partition located on a shared disk (a disk accessible from both master-eligible nodes alternatively, also known as a dual-hosted disk). You can only define a shared partition using a volume manager, for example, Solaris Volume Manager (SVM) on the Solaris OS or Logical Volume Manager (LVM) on Linux. For a listing of all other partition types, see also, partition.
The failure of a master-eligible node, the failure of a service, or the failure of one of the redundant networks. See also, double fault.
If a volume manager is not used, the terms slice and partition refer interchangeably to a physical division of a disk. If a volume manager is used, a slice is equivalent to a metadevice that allows access to a file system. In this case, it can refer either to soft or mirror partitions. See also, partition, soft partition, mirror partition, and metadevice.
A virtual division of one or more disks. You can only define a soft partition using a volume manager, for example, SVM on the Solaris OS or LVM on a Linux distribution. See also, metadevice and slice. For a listing of all other partition types, see also, partition.
An error scenario in which a peer node does not receive information from the master node for more than 10 seconds.
CGTP running on nonpeer nodes. See also, Carrier Grade Transport Protocol and nonpeer node.
A script that launches the Netra HA Suite daemons, some operating system daemons, and some companion product daemons. See also, probe mechanism and nametag.
Part of the AVS software suite, SNDR is used to provide IP replication on the Solaris OS. See also, Distributed Replicated Block Device, IP replication, Reliable File Service.
The planned transfer of a service from the master node to the vice-master node. A switchover can be initiated for the purpose of system recovery or system administration. A switchover is not linked to node failure. See also, failover.
The copying of data from the master node to the vice-master node after the vice-master node has been out of service. Synchronization occurs after startup, disk change, failover, or switchover, unless you have delayed the start of synchronization by setting the RNFS_EnableSync parameter to FALSE in the nhfs.conf file. See also, IP replication.
A state where the replicated partitions on the disks of the master node and vice-master node contain exactly the same data.
A console access device that connects the console ports of several nodes to a TCP/IP network.
After a disk change, cluster startup, failover, or switchover, the master node and vice-master node disks are unsynchronized. That is, the replicated partitions do not contain the same data. See also, synchronization.
The ability to add processors and memory to individual nodes to increase their capability. See also horizontal scalability.
A peer node (master-eligible) that has a copy of all the information on the master node. The vice-master node is available to become the master node in the event of a failover or switchover. See also, master node, master-eligible node, master-ineligible node. For a listing of all other node types, see also, node.
VLANs are software defined groups of hosts on a local area network (LAN) that communicate as if they were on the same wire, even though they are physically on different LAN segments throughout a site.
A logical interface associated with a virtual physical interface such as the CGTP interface. A logical interface associated to the CGTP interface can be called cgtp0:1. One virtual physical interface can have many virtual logical interfaces. See also, virtual physical interface and CGTP interface.
The CGTP interface. The CGTP interface of a node can be called cgtp0. See also virtual logical interface.
One of the Linux distributions provided by Wind River, also called Wind River CGL (Carrier Grade Linux). For more information, see http://www.windriver.com/products/platforms/network_equipment/
| Netra™ High Availability Suite 3.0 1/08 Foundation Services Glossary | 819-5238-13 |
|
Copyright © 2008, Sun Microsystems, Inc. All rights reserved.