Seventh Chapter Equipment Management

first, the hierarchical relationship of equipment

The AIX system divides the hierarchy of the devices into three layers,

1) The top layer is functional classification (functional class), is based on the basic function of the device grouping, indicating what the device can do, for example, all the adapters belong to the adapter class, adapter is a functional classification; All printers belong to the printer functional category All buses belong to bus function classification; All tape drives belong to the tape functional category.

2) The second layer is a functional subclass (functional subclasses), which is divided by a specific property, indicating how the device is connected to the system. Dividing a functional class into multiple functional subclasses, for example, ISA's standard input-output adapter belongs to the Isa_sio function subclass, while the PCI adapter belongs to the PCI function subclass, as if the printer has a serial port and the same port, the serial printer belongs to the RS232 function sub-class, The two-port printer belongs to the parallel function sub-class.

3) The lowest level is the device type, which is categorized by model and manufacturer to indicate what the device is, for example, a tape drive that belongs to the SCSI feature subclass can be divided into 8mm and 4mm two models.

Typically, the predefined device database provides fairly complete support for internal devices, while support for external devices is "on-demand". The user-defined device database contains devices that are currently in existence and are defined and configured in the system, and each device currently connected to the system is a record that is stored in the user-defined device database.

second, logical equipment files

The application is a logical device to operate the physical device, the logical device is the application operation of the physical device interface, the application of data from the logical device read into the memory for processing, after processing, the results are written to the logical device, the operating system to handle the logical device and physical device specific data transmission.

When a user program reads data from a logical device, the operating system passes the device driver's live data from the physical device to the logical device, and when the application writes data to the logical device, the operating system sends the data received on the logical device to the physical device through the device driver.

In UNIX systems, logical devices are often represented as device files, and device files are special files that exist in the/dev directory where users can read and write to logical devices by directly accessing these device folders. The user program accesses the device through a device file, but the read and write operation of the actual device is done by the device driver.

There are two types of devices in Unix systems:

1) block device, random access device with certain structure, such as hard disk, file system, etc.

2) Character devices (Character device), which is a sequential data flow device in which read and write is performed by character, and these characters are continuously formed into a stream of data. It does not have buffers and reads and writes to such devices in real time, such as terminals, tape drives, and so on.

three, the state of the equipment (total 4 kinds)

1) Undefined (undefined): Indicates that the device is not recognized by the system, which means that the device cannot be found in the system. In fact, this state does not exist, only conceptually for a predefined device, there is such a state .

2) Defined (defined): Indicates that the device is recognized by the system, and the information for that device is also stored in the user-defined device database, but it cannot be used by the system. In this state of the device, the system has assigned a logical device name and a device port, and stored the associated properties, but the device cannot be used by the system, the user can not access it.

3) Available (available): Indicates that the device can be used in the system and can be accessed by the user.

4) Stopped (inactive): Indicates that the device cannot be used, but the operating system can identify it through the driver

Iv. machine types for RS6000 and pseries servers

View by #bootinfo-p

Identity

Meaning

rs6k

Micro-channel single processor type

Rs6ksmp

Micro-Channel Multiprocessor type

Rspc

ISA bus Type

Chrp

PCI bus Type

v. Location number of the equipment

The location number is divided into two types, one is the physical location number (physical), which describes the connection of the physical device, and one is the AIX system location number, which describes the logical connection of the physical device in the operating system. Different machine types, their location coding is also different. But the format is the same.

The location number format of the AIX system is differentiated by whether it is a SCSI device, and is divided into two types: non-SCSI device Location number format:ab-cd-ef-gh, SCSI device Location number format:ab-cd-ef-g,h. For both devices, the ab-cd-ef meaning of the position number is the same, except for the last two characters.

1)AB: Indicates the bus type, which is the system board on which the adapter resides. AB is 00 o'clock, indicating that the adapter is in the CPU board or system unit (due to different machine types). When AB is a different value, it means that the adapter is in the I/O expansion board, so AB refers to the I/O bus and the slots on the CPU Board card. Where a represents the I/O bus, which is the bus type, a is 0 o'clock, represents the standard I/O bus, and A is 1 o'clock, which represents the selectable bus. b Specifies the slot number on the I/O bus, which is the position on the bus.

2)CD: Indicates the slot, adapter number, where C represents the I/O motherboard that contains the adapter. If the adapter is in the CPU board or system unit, C is 0 o'clock, it represents the standard I/O bus, and C is 1 o'clock, which represents the selectable I/O bus. If the adapter is in the extended I/O board, C is also 0. D represents the slot number of the adapter in the standard I/O bus (or the slot number in the extended I/O bus).

For PCI-based RS6000, the C in the CD represents the bus type, when C is 0 o'clock, represents the PCI bus, and when C is 1 o'clock, ISA bus. d Specifies the slot number, and D is 1 o'clock, which represents the first specified adapter, which is integrated with SCSI. D is 2 o'clock, indicates a low-end PCI slot, D is 3 o'clock, represents a high-end PCI slot, and for an integrated adapter, D is always 0.

Any one adapter is defined with AB-CD. For SSAS (serial storage fabric) adapters, it is the identity of the system I/O bus.

Therefore, a 00-00 location number is used to represent a standard I/O board in a system unit or CPU board. Below is an example:

00-00: Standard I/O boards in system units or CPU board sinks.

00-06: The adapter is in the 6th slot on the standard I/O board, and the standard motherboard is in the system unit on the CPU board (varies by machine type)

00-15: Identified in the CPU Board card, the adapter is located in the 5th slot of the selectable I/O bus.

3)EF: Identifies the connector number, there may be a connector on the adapter, and EF is used to identify the connector connected to the device. For SSAS devices, EF represents both the physical hard drive module and the logical hard drive module. For PCI-based RS6000, when EF is Od,om,ok, it represents the connector for the disk, mouse, and keyboard, respectively.

For SCSI adapters, the location number is indicated in 00-00-XX format, indicating that the SCSI controller is integrated on the system board without a slot number.

4)gh, g,h: For non-SCSI devices, GH represents a port number, address, DIMM, and Fru, and he has multiple meanings. For SCSI devices, the G in "G,h" indicates the device's control unit number, which is the logical unit number (LUN) of the SCSI Id,h table device, for example: 00-00-s1-00 and 00-00-s2-00 are terminal devices connected to the standard I/O serial ports S1 and S2.

six, the installation of PCI equipment

When using a PCI adapter, the slot number in which it resides is as important as the slot number in the Microchannel device, and do not move the socket of the adapter arbitrarily. If Configuration Manager cannot automatically configure a new device or if the user wants to change a new device property value, the new device must be configured manually.

Two commands below to manage hot-swappable PCI slots in AIX systems

1) #lsslot: Displays a list of slots and their properties

2) #drslot: Reconfigure the slots dynamically.

In addition, you can manage hot-swappable PCI slots with Smit tools or a Web-based system Manager.

To add a hot-swappable PCI adapter, first use the Drslot command to set the status of the slot to the identify (identification) state to verify that the selected slot is correct. After you press ENTER, the slot status changes to the action state. Then install the adapter into the machine and, when finished, press ENTER to open the slot power. This way, the adapter is installed and can be identified with Cfgmgr.

Before removing the adapters from the system, they must be reconfigured in the AIX system, and the adapter must be in a defined (Defined) state or removed from the ODM database.

vii. Multi-Path I/O (MPIO)

AIX5.2 provides multipath I/O (MPIO) technology that allows a single device (such as a disk, LUN) to connect to an adapter that has multiple I/O paths. MPIO is part of the basic kernel that must be located on the same machine or on the same logical partition (LPAR). However, if more than one machine is connected to the same device, this is a cluster rather than multipath I/O (MPIO).

Benefits of MPIO:

1) Improve equipment performance

2) Improve the reliability and availability of equipment

3) make system management easier

The path management feature consists of two modules: Kernel extensions (Pcmke) and runtime loadable configuration modules (PCMRTL).

In a multi-path I/O subsystem, any device may have one or more paths to access it. The path selection for the Pcmke is to detect the path and pass information to the device driver based on the device's configuration. Each MPIO-capable device driver can add a device's path, which is from the device's immediate parent device to itself. When an I/O request is sent to the device, the device driver must decide which path to use to transfer the request. Pcmke chooses the best path according to many criteria: load balancing, connection rate, line failure condition, etc.

Through the modules provided by the device vendor, it is possible to access individual devices through different adapters, but not using MPIO technology. subsystem device drivers like the ESS (IBM storage) (Subsystem devices Driver, or SDD), but need to create a hdisk logical device for each path, while the SDD is responsible for path management.

Path management is handled by the SDD (without the use of MPIO technology), and there are drawbacks:

1) They are dependent on the firmware

2) Differentiated management must be used for each subsystem

3) in ODM, each path corresponds to a logical device record

4) You must use a dedicated command to set up an operating system volume group

If you use MPIO technology, an MPIO device or hdisk can have multiple paths to its parent device (adapter), while the device has only one record in the ODM database. This makes it possible to manage a volume group that contains a MPIO device, which can be managed using a normal AIX command, which has been modified in the AIX device driver in order to achieve this new feature.

unique Device identifier (UDID)

Each MPIO-capable device has a unique identifier, and the MPIO device uses this unique identifier to differentiate all other devices in the system. The Udid value of a specific device can be stored as a device attribute in the ODM device configuration database.

When the Cfgmgr runtime or the configuration placement method of a parent device executes, it requests udid for the child device. The Udid is then compared to the udid stored in the ODM database to determine what action to take:

1) need to define a newly discovered device in the system

2) device already exists, only need to define a path

device pre-defined policies

1. No_reserve in this mode, the MPIO path algorithm enables the I/O to transmit on a single path (Fail_over fail over mode) and the I/O distribution on multipath (load Balancing mode). It is best to use a cluster in the concurrency mode (Concurrent modes). Should not be used in clusters that do not use cluster software (Cluster).

2, Single_path in this mode, the MPIO path algorithm only supports I/O on a single path (Fail_over fail-over mode). It is best to use a cluster that is owned by only one node in the device. If the device owning the node fails, then the standby node takes over this way.

3. Pr_excluster in this mode, the MPIO path algorithm enables the I/O to transmit on a single path (Fail_over fail over mode) and the I/O distribution on multipath (load Balancing mode). It is best to use a cluster that is owned by only one node in the device. If the device owning the node fails, then the standby node takes over this way.

4. pr_shared in this mode, the MPIO path algorithm can support I/O routing on a single path (Fail_over fail over mode) and I/O distribution on multipath (load Balancing mode). This approach is best used in clusters where the device is owned by only one or more nodes.

detects a device that has MPIO capability

In order to detect a SCSI device as a MPIO capable device, additional attributes need to be added to the ODM pre-defined device database (Pdat). All devices that have MPIO capability require Udid,udid to be an attribute in the ODM.

Also, in the ODM pre-defined device database (PDAT), you need to add a PCM (Path control module, multipath Controller) attribute, the PCM attribute points to an ODM helper, which defines the Pcmke module, This module provides path control management capabilities to the drivers for MPIO devices. The PCM attribute contains the name of the Pcmke, which can be provided by the vendor or by the AIX system.

The following SCSI disk has only one path, but it is considered to be a MPIO capable device--

MPIO command Related: Lspath, Mkpath, Chpath, Rmpath

Seventh Chapter Equipment Management