UNIX%20System%20Administration - PowerPoint PPT Presentation

About This Presentation
Title:

UNIX%20System%20Administration

Description:

(IDE) ATA helping if you gave me one.... IDE - The idea of incorporating the ... Add entries in /etc/vfstab as required - enable journaling, but be careful! ... – PowerPoint PPT presentation

Number of Views:53
Avg rating:3.0/5.0
Slides: 19
Provided by: BC860
Learn more at: https://owd.tcnj.edu
Category:

less

Transcript and Presenter's Notes

Title: UNIX%20System%20Administration


1
UNIX System Administration
  • Adding a Disk
  • Chapter 8

2
Disk Interfaces
  • Proprietary interfaces
  • Big. Yellow. Different.
  • SCSI
  • Small Computer Standardized Interface
  • IDE
  • Integrated Drive Electronics
  • Fibre Channel
  • NKOTB

3
(IDE) ATA helping if you gave me one.
  • IDE - The idea of incorporating the drives
    controller (NOT the bus controller) onto the
    drive unit itself.
  • PC IDE is attached to main system bus via only
    buffering/isolation components - there is no IDE
    interface per-se
  • More properly referred to as ATA - Advanced
    Technology Attachment
  • Defined protocol for DISK DRIVES
  • ATAPI - ATA Packet Interface - protocol
    encapsulated in ATA for other IDE devices (CD-?
    Drives, misc removable storage devices)

4
ATA Performance
  • Limited to two devices per channel, two channels
    per system
  • Cheap, fast, single-user storage
  • ATA protocols have less overhead than SCSI
  • No error correction (until ATA-4)
  • No behind-the-scenes management (e.g. integrity
    checks)
  • Less extensible than SCSI
  • Very simple protocol will not support more than
    two drives per channel
  • Drives are not multitasking
  • ATA bus is not shared with neighbor device -
    each drive is master of its own domain while
    handling an I/O request
  • Host CPU must manage drives
  • 60-100 CPU time of a PIII might be spent on a
    large transfer

5
SCSI
  • SCSI - Small Computer Sstandardized Interface
  • Began as Shugart Associates Standardized
    Interface in 1979, adopted as ANSI standard
    (SCSI-1) in 1986 with a name change
  • Separate, intelligent I/O bus attached to main
    bus. All devices on bus are intelligent as well.
  • Managing device - initiator - typically is the
    host interface though this is not required.
    Peripherals are targets.

6
SCSI too
  • Initiator controls whether a SCSI device is
    connected to the bus (separate control lines
    facilitate initiator/target control
    communications)
  • Multiple I/O requests may be issued against
    multiple targets
  • Controllers typically use DMA (bus mastering) to
    offload the work of transferring data to memory
    from the host CPU

7
SCSI Buses
  • Single Ended (SE)
  • Oldest variety
  • Most susceptible to noise due to low signal
    levels - required strict adherence to cable
    specifications
  • Uses common ground for all signal lines
  • Differential varieties
  • Developed to overcome SE limitations - noise
    immunity
  • Uses a and - line set for each signal
  • High Voltage Differential (HVD) - 12 volts
  • Low Voltage Differential (LVD) - 5 volts
  • Less expensive version of HVD
  • LVD devices will detect connection to a SE bus
    and fall back to SCSI-2 speeds
  • All are incompatible
  • All require termination. SE - passive H/LVD -
    active
  • Absorbs signal at ends of the bus to prevent
    reflections

8
Feeling a little...irregular? Add some fibre to
your diet.
  • Fibre Channel - Member of a class of high-speed,
    serial interfaces for disk storage (serial
    referring to the transmission order of data
    bytes)
  • SSA (Serial Storage Architecture) is IBMs
    proprietary attempt at a serial interface for
    storage. As always, Big Blue sticks to that
    P-word. Regardless, SSA is too little, too late.
    Revel in the irony.
  • Fibre Channel media
  • Coaxial copper cable
  • Twisted Pair copper cable
  • Multimode or Singlemode fiber (required for
    speeds over 1Gb/s 100MB/s) - two fibers
    required (Tx/Rx) in a P2P link configuration.
  • Fibre channel protocols
  • Provide capability to support FC-specific
    protocols as well as SCSI, IP, and IEEE 802.2
    (ethernet)
  • Supports class of service - datagram,
    fractional speed, connectionless, etc.

9
FC Topo- and Termi- nology
  • Point-to-point
  • exactly two devices. Exclusive media access.
  • Arbitrated loop
  • up to 127 devices, chained in a loop. Shared
    media access.
  • Fabric-switched
  • up to 224 devices in a cross-point switched
    configuration. Requires a switch, provides
    exclusive media access.

10
Drive Terminology
  • Sector - atomic unit of storage, typically 512B
  • Block - abstracted sector free from track/surface
    designations
  • Track - sectors in a single groove on one surface
    of a platter
  • Cylinder - from a vertical perspective, set of
    tracks on all disk surfaces at equal distance
    from the center
  • Partition - Logical grouping of a set of
    consecutive cylinders or blocks
  • LUN - Logical Unit Number - SCSI construct that
    permits dividing a single target into up to 7
    logical subdevices. All targets have LUN 0,
    devices like multi-tray CD changers and drive
    arrays utilize other LUN values as required.

11
(No Transcript)
12
Preparing your disk
  • Connect the disk to the computer
  • Create device files so disk can be accessed
  • Format the disk
  • Partition the disk
  • Create filesystems on partitions
  • Configure filesystem mounting

13
Let there be light!
  • Connect the disk
  • Create device files to access the disk
  • All devices have entries in the /dev structure
  • Disks (and like storage media) have two
  • Character (or raw) - a low-level device used
    for maintenance interaction (fsck, etc) -
    /dev/rdsk
  • Block - conceptually higher-level device used
    for filesystem interaction (backups, mounting,
    etc) - /dev/dsk
  • Typically established by a reconfiguration
    sequence (though can be established manually
    under certain circumstances)
  • All /dev entries for LUN 0 will be created
    initially, extra LUNs may require extra steps
  • Format the disk
  • In terms of low-level format, disks are
    pre-formatted by the manufacturer. This procedure
    should not be necessary, and should only be
    attempted with sufficient knowledge of the
    ramifications.

14
Click to add title
  • Partition the disk
  • Actual method is O/S dependent
  • Logically groups cylinders (or blocks) into a
    usable portion which can be referenced
  • Partitions are associated with and referred to by
    a device entry
  • CwTxDySz
  • C controller (w) - number of SCSI controller in
    the system
  • T target (x) - SCSI ID of target on bus. 0-7 or
    0-15
  • D LUN (y) - 0 always exists, typically only one
    LUN. 0-7
  • S slice (z) - partition ID. 0-7
  • Slice 2 is the backup slice - a special
    partition that refers to the entire disk, used
    for low-level disk manipulation
  • Partitions should not overlap (but the customer
    is always right)
  • A disks label refers to its partition table map.
    Labeling the disk writes this map onto the disk
    to be used by the system for interfacing with
    that drive.
  • Remember to partition for O/S, swap, and other
    file storage needs (e.g. home, log, application
    directories)

15
Click harder!
  • Create a filesystem on new partitions
  • Partitions (slices) need to be initialized to
    accept a filesystem.
  • Actual procedure varies by O/S - typically the
    newfs command
  • Requires the /dev/rdsk filename of the partition
    to initialize
  • Writes out the required filesystem data
    structures to disk (superblocks, disk block maps,
    inode tables, etc)

16
NOT THAT HARD!
  • Configure filesystem mounting
  • Procedure is (AGAIN) O/S dependent
  • Filesystem must be mounted before can be
    accessible
  • Add entries for automatic mounting
  • /etc/vfstab
  • Mount points are directories on an existing
    filesystem (or / if this is the root disk), but
    once mounted, the new filesystem will overlay and
    hide anything contained in that mount point
    directory.
  • You must mount in order of dependence if mounting
    a filesystem on a mounted filesystem other than /

17
When good disks go bad
  • fsck - Filesystem check and repair
  • Capable of repairing most soft errors
  • Runs automatically on boot (per settings in
    /etc/vfstab)
  • Required to be run against filesystems that were
    potentially made inconsistent by an improper
    shutdown - mount will check a flag that indicates
    the possible state of the filesystem and will
    usually refuse to accept inconsistent filesystems
  • A fsck should only be done on an unmounted
    filesystem especially if changes will be made -
    O/S buffers will become inconsistent and the
    system may panic
  • Certain repair operations may be destructive (but
    would be required before disk could be mounted
    again)
  • Running fsck on a journaled filesystem takes
    little time as the journal log clears up
    inconsistencies (fsck will flush the log, which
    is very small)

18
Solaris
  • Connect the disk (power off, please) and set the
    SCSI Ids and place terminators (or IDE
    master/slave jumper)
  • Power on and enter PROM mode
  • Issue probe-scsi (or probe-scsi-all) to verify
    disk presence
  • Use probe-ide for IDE machines
  • Issue boot -r for reconfig boot
  • Use the format command to label (partition) the
    disk
  • Ensure your slices do not overlap - no direct
    indication would be provided of this oversight
    (the overlap exception is slice 2 - the backup
    slice - dont ever touch it). Zero-out any unused
    partitions to avoid accidental use. Dont forget
    to issue the label command before exiting.
  • Use newfs to initialize the filesystem
  • Mount the filesystem to verify
  • Add entries in /etc/vfstab as required - enable
    journaling, but be careful!!
Write a Comment
User Comments (0)
About PowerShow.com