vi3.5 パフォーマンス esxtop

ストレージネックやストレージのパフォーマンス性能を確認するのに便利なVMwareコマンドがesxtopです。
その他、パフォーマンス状況がわかり易い(VIクライアントのパフォーマンスでも確認できる)CPUやメモリやネットワークといったリソース状況もわかります。

VMwareでストレージのパフォーマンス状況を確認するには、ESXサーバーにログインしrootユーザーになります。
esxtopを実行し、「s」キーを入力します。「s」キーはesxtopの画面更新秒数変更オプションなので「2」を入力し最短の2秒更新画面に変更します。

次に、「u」キーを入力します。「u」キーはESXサーバーのディスクデバイスパフォーマンス状況を確認できる画面への切り替えオプションです。
ESXのディスクパフォーマンス状況で確認すべき項目は
DEVICE
ESXサーバー内のどのディスク(ストレージ)デバイスを確認するか。
QUED
ESXサーバーVMkernelでのデバイスに対するI/Oコマンド数。
このディスクキューがたまっている状態(数がある状態)は、ディスクI/Oキュー待ちが発生している状態。
%USD
ディスクビジー率。(ディスクI/Oキューによるのディスクビジー率)
QUEDに数がある状態や%USDが100%になる状態は、ストレージ(ディスク)ネックになりうる状態です。

esxtop help


Esxtop version 3.1.0
Secure mode Off

Esxtop: top for ESX

These single-character commands are available:

^L      - redraw screen
space   - update display
h or ?  - help; show this text
q       - quit

Interactive commands are:

fF      Add or remove fields
oO      Change the order of displayed fields
s       Set the delay in seconds between updates
#       Set the number of instances to display
W       Write configuration file ~/.esxtop3rc
e       Expand/Rollup Cpu Statistics
V       View only VM instances

Sort by:
        U:%USED         R:%RDY          N:Default
Switch display:
        m:ESX memory    d:ESX disk adapter      u:ESX disk device       v:ESX disk VM   n:ESX nic
        c:ESX cpu

ESX cpu

Current Field order: ABCDEfgh

* A:  ID = Id
* B:  GID = Group Id
* C:  NAME = Name
* D:  NWLD = Num Members
* E:  %STATE TIMES = CPU State Times
  F:  EVENT COUNTS/s = CPU Event Counts
  G:  CPU ALLOC = CPU Allocations
  H:  SUMMARY STATS = CPU Summary Stats


ESX memory

Current Field order: aBCDefGHijklMno

  A:  ID = Id
* B:  GID = Group Id
* C:  NAME = Name
* D:  NWLD = Num Members
  E:  MEM ALLOC = MEM Allocations
  F:  NUMA STATS = Numa Statistics
* G:  SIZE = MEM Size (MB)
* H:  ACTV = MEM Active (MB)
  I:  MCTL = MEM Ctl (MB)
  J:  SWAP STATS = Swap Statistics (MB)
  K:  CPT = MEM Checkpoint (MB)
  L:  COW = MEM Cow (MB)
* M:  OVHD = MEM Overhead (MB)
  N:  CMT = MEM Committed (MB)
  O:  RESP? = MEM Responsive?

ESX disk adapter

Current Field order: ABCDEfgHIjklm

* A:  ADAPTR = Adapter Name
* B:  LUNID = Lun Id
* C:  WID = World Id
* D:  NLUNS = Num Luns
* E:  NVMS = Num VMs
  F:  SHARES = Shares
  G:  BLKSZ = Block Size (bytes)
* H:  QSTATS = Queue Stats
* I:  IOSTATS = I/O Stats
  J:  LATSTATS = Latency stats (ms)
  K:  ERRSTATS/s = Error Stats
  L:  PAESTATS/s = PAE Stats
  M:  SPLTSTATS/s = SPLIT Stats


ESX disk device

Current Field order: ABCdeFGhijk

* A:  DEVICE = Device Name
* B:  ID = Path/World/Partition Id
* C:  NUM = Num of Objects
  D:  SHARES = Shares
  E:  BLKSZ = Block Size (bytes)
* F:  QSTATS = Queue Stats
* G:  IOSTATS = I/O Stats
  H:  LATSTATS = Latency stats (ms)
  I:  ERRSTATS/s = Error Stats
  J:  PAESTATS/s = PAE Stats
  K:  SPLTSTATS/s = SPLIT Stats


ESX disk VM

Current Field order: ABCDEfgHIjk

* A:  ID = World Id
* B:  GID = Grp Id
* C:  NAME = Group/World Name
* D:  DEVICE = Device Name
* E:  NUM = Num of Objects
  F:  SHARES = Shares
  G:  BLKSZ = Block Size (bytes)
* H:  QSTATS = Queue Stats
* I:  IOSTATS = I/O Stats
  J:  LATSTATS = Latency stats (ms)
  K:  ERRSTATS/s = Error Stats


ESX nic

Current Field order: ABcDEFGHIJKLm

* A:  PORT ID = Port Id
* B:  UPLINK = Uplink(Y/N)
  C:  PNIC = Physical Nic Properties
* D:  USED BY = Used By Name
* E:  DTYP = Device Type
* F:  DNAME = Device Name
* G:  PKTTX/s = Packets Tx/s
* H:  MbTX/s = MegaBits Tx/s
* I:  PKTRX/s = Packets Rx/s
* J:  MbRX/s = MegaBits Rx/s
* K:  DRPTX/s = %Packets Dropped (Tx)
* L:  DRPRX/s = %Packets Dropped (Rx)
  M:  ACTN/s = Actions/s


man esxtop


ESXTOP(1)                  VMware ESX Server Manual                  ESXTOP(1)



NAME
       esxtop, resxtop - display ESX Server resource utilization statistics

COPYRIGHT
       VMware  ESX  Server  is  Copyright  2006-2007  VMware, Inc.  All rights
       reserved.

SYNOPSIS
       esxtop [-] [h] [v] [b] [s] [a] [c configfile]  [R  vm-support_dir_path]
       [d delay] [n iter]

       resxtop  [-] [h] [v] [b] [s] [a] [c configfile] [R vm-support_dir_path]
       [d  delay]  [n  iter]  [-server  host-name]  [-portnumber  socket-port]
       [-username user-name]

DESCRIPTION
       esxtop provides a fine-grain look at resource utilization in real time.
       For more information,  please  refer  to  the  official  documentation,
       available  at  <http://www.vmware.com/info?id=193>.  Esxtop runs on the
       ESX Server Service Console. It can only be run by user root.  It can be
       run  in three different modes; interactive (default), batch and replay.
       Worlds in this document refer to ESX Server VMKernel schedulable  enti-
       ties, similar to processes in other operating systems.

       resxtop  provides a fine-grain look at resource utilization of a speci-
       fied host in real time. Resxtop connects to the host using a user  name
       and password. Options --server, --portnumber and --username can be used
       to specify the connection parameters. The resxtop user will be prompted
       to  input  the password. Resxtop can be run in three different modes in
       the same way as esxtop; interactive (default), batch and replay.

Interactive Mode
       This is the default mode for  esxtop.  This  mode  displays  statistics
       under  four  broad categories. Each category is displayed on a seperate
       screen.  There are screens displaying CPU, memory, storage and  network
       resource usage.  A help menu is available for each of the four screens.
        In this mode there are several command line options available.

   Command-line Options
       h    Prints help for esxtop command-line options.

       v    Prints esxtop version number.

       s    Tells esxtop to run in secure mode.  This disables the interactive
            command to change delay between screen updates.

       a    Show  all statistics. This option overides ~/.esxtop310rc configu-
            ration file setups and shows all statistics.

       c    Load user-defined configuration file. Default  configuration  file
            is  ~/.esxtop310rc.   User  can specify the filename to be used as
            configuration  file.   User-defined  configuration  file  will  be
            default file name for saving configuration with W command.

       d    Specifies the delay between screen updates.  Default is 5 seconds.
            Minimum is 2 seconds.  You can change this  with  the  interactive
            command  's'.  If a delay of less than 2 seconds is specified then
            the delay is set to 2 seconds.

       n    Number of iterations.  Update the display 'n' number of times  and
            then exit.

   Common Statistics Description
       Several  statistics are displayed on the different screens while esxtop
       is running in interactive mode. Statistics listed in this  section  are
       common across all four screens.

       uptime
            This  first  line  found  at  the  top  of each of the four esxtop
            screens displays the current time, time since last reboot and num-
            ber of currently running Worlds.
             On  this  first line CPU load averages are also displayed for the
            CPU, storage and network screens.  The load averages over the past
            1, 5 and 15 minutes are displayed. Load averages take into account
            both running and ready-to-run  Worlds.  A  load  average  of  1.00
            implies that all the physical CPUs are fully accounted for. A load
            average of 2.00 implies that the ESX Server  may  be  in  need  of
            twice as many physical CPUs as currently available.
             For  the  memory  screen memory overcommit averages are also dis-
            played on the first line.  The memory overcommit averages over the
            past  1, 5 and 15 minutes are displayed. A memory overcommit aver-
            age of 0.37 implies that the memory is overcommit by 37%.

   Common Interactive Commands
       Several single-key commands are recognized while esxtop is  running  in
       interactive mode. Commands listed in this section are common across all
       four screens. The command to specify the delay between updates is  dis-
       abled  if  the s option has been given on the command line. All sorting
       interactive commands sort in descending order.

       h or ?
            Displays a help menu for the current screen giving a brief summary
            of commands, and the status of secure mode.

       space
            Immediately updates the current screen.

       ^L   Erases and redraws the current screen.

       f or F
            Displays  a screen to add or remove statistics columns (fields) to
            or from the current screen. See below for more information.

       o or O
            Displays a screen to change order of  statistics  columns  on  the
            current screen.  See below for more information.

       #    Change the number of statistics rows to show. You will be prompted
            to enter the number. Any value greater than 0 overrides  automatic
            determination  of  the  number  of rows to show, which is based on
            window size measurement. This number updated on  any  one  of  the
            four esxtop screens will be used for all esxtop screens.

       s    Change  the  delay between updates.  You will be prompted to enter
            the delay time, in seconds, between updates. Fractional values are
            recognized  down  to microseconds. The default value is 5 seconds.
            The minimum value is 2 seconds. This command is not  available  in
            secure mode.

       W    Write current setup to ~/.esxtop310rc, or to user-defined configu-
            ration file specified by -c option.  This is the  recommended  way
            to write an esxtop configuration file. See below for more informa-
            tion.

       q    Quit.

       c    switch to CPU resource utilization screen.

       m    switch to memory resource utilization screen.

       d    switch to storage (disk) adapter resource utilization screen.

       u    switch to storage (disk) device resource utilization screen.

       v    switch to storage (disk) VM resource utilization screen.

       n    switch to network resource utilization screen.

   The Statistics Columns (fields) and Order Screens
       After pressing f, F, o or O you will be shown a screen  specifying  the
       field  order  on  the top line and short descriptions of the field con-
       tents. The field order string uses the following syntax: if the  letter
       in  the field string corresponding to a  field is upper case, the field
       will be displayed.  This is also indicated by an asterisk in  front  of
       the  field  description.   The  order  of the fields corresponds to the
       order of the letters in the string.
        From the field select screen you can toggle the display of a field  by
       pressing the corresponding letter.
        From the order screen you may move a field to the left by pressing the
       corresponding upper case letter and to the right by pressing the  lower
       case one.

   Configuration File
       esxtop  reads  its  default  configuration  from ~/.esxtop310rc or from
       specified file name with -c option.  This configuration  file  contains
       five  lines.  The first four lines contain lower and upper case letters
       to specify which fields in what order are to be displayed on  the  CPU,
       memory, storage and network screens. The letters correspond to the let-
       ters in the Fields or Order screens for the respective esxtop  screens.
       Editing  this  file by hand is not recommended. Select fields and order
       in a running esxtop process and save this file using the W  interactive
       command.   The  fifth  line  contains information on the other options.
       Most important, if you have saved a configuration in secure  mode,  you
       will  not  get  an  insecure esxtop without removing the lower 's' from
       this fifth line of your ~/.esxtop310rc.  A digit  specifies  the  delay
       time  between  updates.   As in interactive mode, a lower 'c', 'm', 'd'
       and 'n' determines the screen esxtop starts on.

   CPU Screen
       This screen displays server-wide and per Resource Pool/World  CPU  uti-
       lization  statistics.  By default, World CPU utilization statistics are
       aggregated into Resource Pools of Worlds. For  Worlds  belonging  to  a
       Virtual  Machine,  these  Resource  Pools correspond to running Virtual
       Machines. All other Worlds are logically aggregated into Resource Pools
       to save vertical window space.

       Statistics Descriptions

       PCPU(%)
            This line displays the percentage CPU utilization per physical CPU
            and the total average physical CPU utilization.

       LCPU(%)
            This line displays the percentage CPU utilization per logical CPU.
            The CPU used percentages for the logical CPUs belonging to a pack-
            age add up to 100%.  This line is displayed only if  hyper-thread-
            ing is present and enabled.

       CCPU(%)
            This  line  displays the percentages of total CPU time as reported
            by the ESX Server Service Console.   us  is  for  percentage  user
            time,  sy is for percentage system time, id is for percentage idle
            time and wa is for percentage wait time.  cs/sec is for  the  con-
            text  switches  per  second  recorded  by  the  ESX Server Service
            Console.

       ID   The Resource Pool ID of the running World's Resource Pool or World
            ID of running World.

       GID  The Resource Pool ID of running World's Resource Pool.

       NAME The  name  of  running  World's  Resource  Pool or name of running
            World.

       NWLD The number of worlds  in  running  World's  Resource  Pool.  If  a
            Resource  Pool  is  expanded  using the interactive command e (see
            interactive commands below) then NWLD for all the resulting Worlds
            belonging  to the Resource Pool will be 1 (some Resource Pool like
            the console Resource Pool have only 1 member).

       %STATE TIMES
            This is a set of CPU statistics made up of the following  percent-
            ages.   Note that for a World, the percentages are a percentage of
            one physical CPU.

       %USED
            The percentage physical CPU used by the Resource Pool/World.

       %SYS The percentage of time spent in the ESX Server VMKernel on  behalf
            of  the  Resource  Pool/World to process interrupts and to perform
            other system activities. This time is part of used  time  used  to
            calculate %USED above.

       %OVRLP
            The  Percentage  of  system  time that was spent on behalf of some
            other Resource Pool/World while Resource Pool/World was scheduled.
            This time is not included in the %SYS for Resource Pool/World. For
            example, if Resource Pool A was currently scheduled and a  network
            packet  for  Resource Pool B was being processed by the ESX Server
            VMKernel, then the time spent  doing  so  appears  as  %OVRLP  for
            Resource Pool A and %SYS for Resource Pool B.

       %RUN Percentage of total time scheduled. This time does not account for
            hyper-threading and  system  time.  Hence,  on  a  hyper-threading
            enabled server, the %RUN can be twice as large as %USED.

       %WAIT
            The total percentage of time the Resource Pool/World spent in wait
            state.  Note that this percentage includes the percentage of  time
            the Resource Pool/World was idle.

       %CSTP
            The percentage of time the Resource Pool/World spent in ready, co-
            deschedule state. This statistic is for  internal  to  VMware  use
            only.

       %IDLE
            The percentage of time the Resource Pool/World was idle. Subtract-
            ing this percentage from %WAIT above gives you the  percentage  of
            time the Resource Pool/World was waiting on some event.

       %RDY The percentage of time the Resource Pool/World was ready to run.

       %MLMTD
            Percentage of time the ESX Server VMKernel deliberately didn't run
            the Resource Pool/World because that would  violate  the  Resource
            Pool/World's limit setting. Even though the Resource Pool/World is
            ready to run when it is prevented from running in  this  way,  the
            %MLMTD time is not included in %RDY time.

       EVENT COUNTS/s
            This is a set of CPU statistics made up of the following, per sec-
            ond, event rates. These statistics are for internal to VMware  use
            only.

       CPU ALLOC
            This is a set of CPU statistics made up of the following CPU allo-
            cation configuration parameters.

       AMIN Resource Pool/World attribute reservation.

       AMAX Resource Pool/World attribute limit.

       ASHRS
            Resource pool attribute shares.

       SUMMARY STATS
            This is a set of CPU statistics made up of the following CPU  con-
            figuration  parameters and statistics.  Note that these statistics
            are only applicable to Worlds and not Resource Pools.

       AFFINITY
            Bit mask showing the current scheduling affinity for the World.

       HTSHARING
            current hyper-threading configuration.

       CPU  The physical or logical processor the World was found to  be  run-
            ning on.

       HTQ  World  is  currently  quaratined  or  not.  'N' implies no and 'Y'
            implies yes.

       TIMER/s
            The timer rate the World is currently requesting for.

       Interactive Commands

       e    Expand/Rollup Resource Pool/Worlds CPU statistics. Allows  viewing
            CPU  resource  utilization  statistics  broken  down by individual
            Worlds belonging to a Resource Pool.

       U    Sort Resource Pools/Worlds by Resource Pool %USED column.

       R    Sort Resource Pools/Worlds by Resource Pool %RDY column.

       N    Sort Resource Pools/Worlds by GID column. This is the default sort
            order.

       V    Display VM instances only.

   Memory Screen
       This  screen displays server-wide and per Resource Pool memory utiliza-
       tion statistics. As on the CPU screen,  Resource  Pools  correspond  to
       running  Virtual  Machines or Worlds that consume memory. Note the dis-
       tinction between machine memory and physical memory in this section.

       Statistics Descriptions

       PMEM (MB)
            This line displays the machine memory statistics for  the  server.
            All  numbers  are  in megabytes.  total is for the total amount of
            machine memory in the server, cos is for  the  amount  of  machine
            memory allocated to the ESX Server Service Console, vmk is for the
            amount of machine memory being used by the  ESX  Server  VMKernel,
            other is for the amount of machine memory being used by everything
            other than the ESX Server Service Console and ESX Server  VMKernel
            and free is for the amount of machine memory that is free.

       VMKMEM (MB)
            This  line  displays  the  machine  memory  statistics for the ESX
            Server VMKernel.  All numbers are in megabytes.   managed  is  for
            the  total amount of machine memory managed by the ESX Server VMK-
            ernel, minfree is for the minimum amount of  machine  memory  that
            the  ESX  Server VMKernel would like to keep free, rsvd is for the
            amount of machine memory that is currently reserved, ursvd is  for
            the  amount  of  machine  memory  that is currently unreserved and
            state is for the memory state as reported by the ESX Server VMKer-
            nel.   Possible  values  are high, soft, hard and low.  Here, high
            implies that the machine memory is not under any pressure and  low
            implies that the machine memory is under pressure.

       COSMEM (MB)
            This  line  displays  the memory statistics as reported by the ESX
            Server Service Console.  All numbers are in  megabytes.   free  is
            for  the  amount  of  idle machine memory, swap_t is for the total
            swap configured, swap_f is for the amount of swap free, r/s is for
            the  rate  at  which memory is swapped in from disk and w/s is for
            the rate at which  memory is swapped to disk.

       NUMA (MB)
            This line displays the ESX Server NUMA statistics.  This  line  is
            only  displayed if the ESX Server is running on a NUMA server. All
            numbers are in megabytes.  For each NUMA node in the server  there
            are  two statistics that are displayed. The first statistic is the
            total amount of machine memory in the NUMA node that is managed by
            the  ESX  Server.  The  second statistic, that is displayed within
            round brackets, is the amount of machine memory in the  node  that
            is currently free.

       PSHARE (MB)
            This  line  displays  the ESX Server page-sharing statistics.  All
            numbers are in megabytes.  shared is for the  amount  of  physical
            memory  that  is being shared, common is for the amount of machine
            memory that is common across World(s) and saving is for the amount
            of machine memory that is saved due to page-sharing.

       SWAP (MB)
            This line displays the ESX Server swap usage statistics.  All num-
            bers are in megabytes.  curr is for the current swap usage, target
            is  for  where the ESX Server expects the swap usage to be, r/s is
            for the rate at which memory is swapped in by the ESX Server  from
            disk and w/s is for the rate at which memory is swapped to disk by
            the ESX Server.

       MEMCTL (MB)
            This line displays the Memory balloon statistics.  All numbers are
            in  megabytes.   curr  is  for the total amount of physical memory
            reclaimed using the vmmemctl modules,  target  is  for  the  total
            amount  of  physical  memory  the ESX Server would like to reclaim
            using the vmmemctl modules and max is for the  maximum  amount  of
            physical memory the ESX Server can reclaim using the vmmemctl mod-
            ules.

       AMIN Resource pool attribute reservation.

       AMAX Resource pool attribute limit.

       ASHRS
            Resource pool attribute shares.

       NHN  Current Home Node for Resource Pool. This statistic is only appli-
            cable on NUMA systems.

       NRMEM (MB)
            Current  amount  of remote memory being accessed by Resource Pool.
            This statistic is only applicable on NUMA systems.

       NLMEM (MB)
            Current amount of local memory being accessed  by  Resource  Pool.
            This statistic is only applicable on NUMA systems.

       N%L  Current  percentage memory being accessed by Resource Pool that is
            local.  This statistic is only applicable on NUMA systems.

       GST_NDx (MB)
            The guest memory being allocated for Resource Pool on NUMA node x.
            "x" is the node number.  This statistic is only applicable on NUMA
            systems.

       OVD_NDx (MB)
            The VMM overhead memory being allocated for Resource Pool on  NUMA
            node  x.  "x" is the node number.  This statistic is only applica-
            ble on NUMA systems.

       MEMSZ (MB)
            The amount of physical memory allocated to a Resource Pool.

       SZTGT (MB)
            The amount of machine memory the  ESX  Server  VMKernel  wants  to
            allocate to Resource Pool.

       TCHD (MB)
            The working set estimate for Resource Pool.

       %ACTV
            Percentage  of  guest  physical memory that is being referenced by
            the guest. This is an instantaneous value.

       %ACTVS
            Percentage of guest physical memory that is  being  referenced  by
            the guest. This is a slow moving average.

       %ACTVF
            Percentage  of  guest  physical memory that is being referenced by
            the guest. This is a fast moving average.

       %ACTVN
            Percentage of guest physical memory that will be referenced by the
            guest.  This  is  an estimation. This statistic is for internal to
            VMware use only.

       MCTL?
            memory balloon driver is installed or not. 'N' implies no and  'Y'
            imples yes.

       MCTLSZ (MB)
            The  amount of physical memory reclaimed from Resource Pool by way
            of ballooning.

       MCTLTGT (MB)
            The amount of physical memory the ESX Server would like to reclaim
            from Resource Pool by way of ballooning.

       MCTLMAX (MB)
            The  maximum  amount of physical memory the ESX Server can reclaim
            from Resource Pool by way of ballooning.  This maximum  is  depen-
            dent on guest operating system type.

       SWCUR (MB)
            Current swap usage by Resource Pool.

       SWTGT (MB)
            Where  the  ESX Server expectd the Swap usage by the Resource Pool
            to be.

       SWR/s (MB)
            Rate at which memory is being swapped in by the  ESX  Server  from
            disk for the Resource Pool.

       SWW/s (MB)
            Rate at which Resource Pool memory is being swapped to disk by the
            ESX Server.

       CPTRD (MB)
            Amount of data read from checkpoint file.

       CPTTGT (MB)
            Size of checkpoint file.

       ZERO (MB)
            Resource Pool physical pages that are zeroed.

       SHRD (MB)
            Resource Pool physical pages that are shared.

       SHRDSVD (MB)
            Machine pages that are saved due to Resource Pool shared pages.

       COWH (MB)
            This statistic is for internal to VMware use only.

       OVHDUW (MB)
            Current space overhead for the user world. This statistic  is  for
            internal to VMware use only.

       OVHD (MB)
            Current space overhead for Resource Pool.

       OVHDMAX (MB)
            Maximum space overhead that may be incurred by Resource Pool.

       MEM Committed (MB)
            This  is a set of memory statistics made up of minimum memory com-
            mitment target, current memory commitment target,  charged  memory
            commitment, and pages per share. These statistics are for internal
            to VMware use only.

       MEM Responsive
            This statistic is for internal to VMware use only.

       Interactive Commands

       V    Display VM instances only.

       M    Sort Resource Pools by Resource Pool MEMSZ column.

       B    Sort Resource Pools by Resource Pool MCTLSZ column.

       N    Sort Resource Pools by GID column. This is the default sort order.

   Storage Adapter Screen
       This  screen  displays  server-wide  storage utilization statistics. On
       this screen statistics are aggregated per storage adapter  by  default.
       Statistics  can  also  be  viewed  per storage channel, target, path or
       world using a LUN. See below for more information.

       Statistics Descriptions

       ADAPTR
            The name of the storage adapter.

       CID  The storage adapter channel id. This id is  only  visible  if  the
            corresponding  adapter  is  expanded.  See interactive command 'e'
            below.

       TID  The storage adapter channel target id. This id is only visible  if
            the  corresponding  adapter and channel are expanded. See interac-
            tive commands 'e' and 'a' below.

       LID  The storage adapter channel target LUN id. This id is only visible
            if the corresponding adapter, channel and target are expanded. See
            interactive commands 'e', 'a' and  't'  below.  ADAPTR:CID:TID:LID
            specifies  one  path.   When  the  statistics  are expanded to lun
            level, esxtop shows the path statistics.

       WID  The storage adapter channel target LUN World id. This id  is  only
            visible  if the corresponding adapter, channel, target and LUN are
            expanded. See interactive commands 'e', 'a', 't'  and  'l'  below.
            The world statistics are per world per device.

       NCHNS
            The number of channels.

       NTGTS
            The number of targets.

       NLUNS
            The number of LUNs.

       NVMS The number of Worlds.

       SHARES
            The  number  of  shares.   This  statistic  is  only applicable to
            worlds.

       BLKSZ
            The block size in bytes. This  statistic  is  only  applicable  to
            LUNs.

       AQLEN
            The  storage  adapter  queue depth.  This is the maximum number of
            ESX Server VMKernel active commands that  the  adapter  driver  is
            configured to support.

       LQLEN
            The  LUN  queue  depth.   This is the maximum number of ESX Server
            VMKernel active commands that the LUN is allowed to have.

       WQLEN
            The World queue depth.  This is the maximum number of  ESX  Server
            VMKernel  active commands that the World is allowed to have.  Note
            that this is a per LUN maximum for the World.

       ACTV The number of commands in the ESX Server VMKernel  that  are  cur-
            rently  active.   This  statistic is only applicable to worlds and
            LUNs.

       QUED The number of commands in the ESX Server VMKernel  that  are  cur-
            rently  queued.   This  statistic is only applicable to worlds and
            LUNs.

       %USD The percentage of queue depth used by ESX Server  VMKernel  active
            commands.  This statistic is only applicable to worlds and LUNs.

       LOAD The  ratio  of ESX Server VMKernel active commands plus ESX Server
            VMKernel queued commands to queue depth.  This statistic  is  only
            applicable to worlds and LUNs.

       CMDS/s
            The  number  of commands issued per second.  If multiple paths are
            connected to the same lun, this statistic is per path when  it  is
            expanded  to luns; it is per world per device, when it is expanded
            to worlds.

       READS/s
            The number of read commands issued per second.  If multiple  paths
            are  connected to the same lun, this statistic is per path when it
            is expanded to luns; it is  per  world  per  device,  when  it  is
            expanded to worlds.

       WRITES/s
            The number of write commands issued per second.  If multiple paths
            are connected to the same lun, this statistic is per path when  it
            is  expanded  to  luns;  it  is  per  world per device, when it is
            expanded to worlds.

       MBREAD/s
            The megabytes read per second.  If multiple paths are connected to
            the  same  lun,  this statistic is per path when it is expanded to
            luns; it is per world per device, when it is expanded to worlds.

       MBWRTN/s
            The megabytes written per second.  If multiple paths are connected
            to the same lun, this statistic is per path when it is expanded to
            luns; it is per world per device, when it is expanded to worlds.

       DAVG/cmd
            The average device latency (millisecs) per command.

       KAVG/cmd
            The average ESX Server VMKernel latency (millisecs) per command.

       GAVG/cmd
            The average Guest OS latency (millisecs) per command.

       QAVG/cmd
            The average queue latency (millisecs) per command.

       DAVG/rd
            The average device read latency (millisecs) per read.

       KAVG/rd
            The average ESX Server VMKernel read latency (millisecs) per read.

       GAVG/rd
            The average Guest OS read latency (millisecs) per read.

       QAVG/rd
            The average queue read latency (millisecs) per read.

       DAVG/wr
            The average device write latency (millisecs) per write.

       KAVG/wr
            The  average  ESX  Server  VMKernel  write latency (millisecs) per
            write.

       GAVG/wr
            The average Guest OS write latency (millisecs) per write.

       QAVG/wr
            The average queue write latency (millisecs) per write.

       ABRTS/s
            The number of commands aborted per second.  If multiple paths  are
            connected  to  the same lun, this statistic is per path when it is
            expanded to luns; it is per world per device, when it is  expanded
            to worlds.

       RESETS/s
            The  number  of  commands reset per second.  If multiple paths are
            connected to the same lun, this statistic is per path when  it  is
            expanded  to luns; it is per world per device, when it is expanded
            to worlds.

       PAECMD/s
            The number of PAE commands per  second.   This  statistic  is  not
            applicable to worlds.

       PAECP/s
            The number of PAE copies per second.  This statistic is not appli-
            cable to worlds.

       SPLTCMD/s
            The number of split commands per second.  This  statistic  is  not
            applicable to worlds.

       SPLTCP/s
            The  number  of  split  copies  per second.  This statistic is not
            applicable to worlds.

   Storage Device Screen
       This screen displays server-wide  storage  utilization  statistics.  On
       this  screen  statistics  are aggregated per storage device by default.
       Statistics can also be viewed per path, world, or partition.  See below
       for more information.

       Statistics Descriptions

       DEVICE
            The name of the storage device.

       PATH The  path  name.  This  name  is only visible if the corresponding
            device is expanded to paths. See interactive command 'p' below.

       WORLD
            The world id. This id is only visible if the corresponding  device
            is  expanded  to  worlds.  See interactive command 'e' below.  The
            world statistics are per world per device.

       PARTITION
            The partition id. This id is only  visible  if  the  corresponding
            device  is  expanded  to  partitions.  See interactive command 't'
            below.

       NPH  The number of paths.

       NWD  The number of worlds.

       NPN  The number of partitions.

       SHARES
            The number of  shares.   This  statistic  is  only  applicable  to
            worlds.

       BLKSZ
            The block size in bytes.

       NUMBLKS
            The number of blocks of the device.

       DQLEN
            The storage device queue depth.  This is the maximum number of ESX
            Server VMKernel active commands that the device is  configured  to
            support.

       WQLEN
            The  World  queue depth.  This is the maximum number of ESX Server
            VMKernel active commands that the World is allowed to have.   Note
            that this is a per device maximum for the World.  It is valid only
            if the corresponding device is expanded to worlds.

       ACTV The number of commands in the ESX Server VMKernel  that  are  cur-
            rently  active.   This  statistic is only applicable to worlds and
            devices.

       QUED The number of commands in the ESX Server VMKernel  that  are  cur-
            rently  queued.   This  statistic is only applicable to worlds and
            devices.

       %USD The percentage of queue depth used by ESX Server  VMKernel  active
            commands.   This  statistic  is  only  applicable  to  worlds  and
            devices.

       LOAD The ratio of ESX Server VMKernel active commands plus  ESX  Server
            VMKernel  queued  commands to queue depth.  This statistic is only
            applicable to worlds and devices.

       CMDS/s
            The number of commands issued per second.

       READS/s
            The number of read commands issued per second.

       WRITES/s
            The number of write commands issued per second.

       MBREAD/s
            The megabytes read per second.

       MBWRTN/s
            The megabytes written per second.

       DAVG/cmd
            The average device latency (millisecs) per command.

       KAVG/cmd
            The average ESX Server VMKernel latency (millisecs) per command.

       GAVG/cmd
            The average Guest OS latency (millisecs) per command.

       QAVG/cmd
            The average queue latency (millisecs) per command.

       DAVG/rd
            The average device read latency (millisecs) per read.

       KAVG/rd
            The average ESX Server VMKernel read latency (millisecs) per read.

       GAVG/rd
            The average Guest OS read latency (millisecs) per read.

       QAVG/rd
            The average queue read latency (millisecs) per read.

       DAVG/wr
            The average device write latency (millisecs) per write.

       KAVG/wr
            The  average  ESX  Server  VMKernel  write latency (millisecs) per
            write.

       GAVG/wr
            The average Guest OS write latency (millisecs) per write.

       QAVG/wr
            The average queue write latency (millisecs) per write.

       ABRTS/s
            The number of commands aborted per second.

       RESETS/s
            The number of commands reset per second.

       PAECMD/s
            The number of PAE commands per second.   This  statistic  is  only
            applicable to paths.

       PAECP/s
            The  number  of  PAE  copies  per  second.  This statistic is only
            applicable to paths.

       SPLTCMD/s
            The number of split commands per second.  This statistic  is  only
            applicable to paths.

       SPLTCP/s
            The  number  of  split  copies per second.  This statistic is only
            applicable to paths.

       Interactive Commands

       e    Expand/Rollup storage world statistics.   Allows  viewing  storage
            resource  utilization  statistics broken down by individual worlds
            belonging to an expanded storage device. You will be  prompted  to
            enter the device name. The statistics are per world per device.

       p    Expand/Rollup  storage  path  statistics.  Allows  viewing storage
            resource utilization statistics broken down  by  individual  paths
            belonging  to  an expanded storage device. You will be prompted to
            enter the device name.

       t    Expand/Rollup storage partition statistics. Allows viewing storage
            resource  utilization  statistics broken down by individual parti-
            tions belonging  to  an  expanded  storage  device.  You  will  be
            prompted to enter the device name.

       r    Sort by READS/s column.

       w    Sort by WRITES/s column.

       R    Sort by MBREAD/s column.

       T    Sort by MBWRTN/s column.

       N    Sort  first by DEVICE column, then by PATH/WORLD/PARTITION column.
            This is the default sort order.

   Storage VM Screen
       This screen displays VM-centric storage  statistics.  On  this  screen,
       statistics  are aggregated on a per-resource-pool basis by default. One
       VM has one corresponding resource pool, so, they are equivalent to per-
       VM statistics.  Statistics can also be viewed on a per-world, or a per-
       world-per-device basis.

       Statistics Descriptions

       ID   The Resource Pool ID of the running World's Resource  Pool  or  ,B
            World ID of runnng World.

       GID  The Resource Pool ID of running World's Resource Pool.

       NAME The  name  of running World's Resource Pool or name of the running
            World.

       Device
            The storage device name. This name is only visible if  correspond-
            ing  world  is  expanded  to  devices. See interactive command 'i'
            below.

       NWD  The number of worlds.

       NDV  The number of devices.  It is  valid  only  if  the  corresponding
            resource pool is expanded to worlds

       SHARES
            The  number  of  shares.   This  statistic  is  only applicable to
            worlds.  It is valid only if the corresponding  resource  pool  is
            expaneded to worlds

       BLKSZ
            The  block  size  in  bytes. It is valid only if the corresponding
            world is expanded to devices.

       NUMBLKS
            The number of blocks of the device. It is valid only if the corre-
            sponding world is expanded to devices.

       DQLEN
            The storage device queue depth.  This is the maximum number of ESX
            Server VMKernel active commands that the device is  configured  to
            support.  It  is valid only if the corresponding world is expanded
            to devices.

       WQLEN
            The World queue depth.  This is the maximum number of  ESX  Server
            VMKernel  active commands that the World is allowed to have. It is
            valid only if the corresponding world is expanded to worlds.

       ACTV The number of commands in the ESX Server VMKernel  that  are  cur-
            rently  active.   This  statistic is only applicable to worlds and
            devices.

       QUED The number of commands in the ESX Server VMKernel  that  are  cur-
            rently  queued.   This  statistic is only applicable to worlds and
            devices.

       %USD The percentage of queue depth used by ESX Server  VMKernel  active
            commands.   This  statistic  is  only  applicable  to  worlds  and
            devices.

       LOAD The ratio of ESX Server VMKernel active commands plus  ESX  Server
            VMKernel  queued  commands to queue depth.  This statistic is only
            applicable to worlds and devices.

       CMDS/s
            The number of commands issued per second.

       READS/s
            The number of read commands issued per second.

       WRITES/s
            The number of write commands issued per second.

       MBREAD/s
            The megabytes read per second.

       MBWRTN/s
            The megabytes written per second.

       DAVG/cmd
            The average device latency (millisecs) per command.

       KAVG/cmd
            The average ESX Server VMKernel latency (millisecs) per command.

       GAVG/cmd
            The average Guest OS latency (millisecs) per command.

       QAVG/cmd
            The average queue latency (millisecs) per command.

       DAVG/rd
            The average device read latency (millisecs) per read.

       KAVG/rd
            The average ESX Server VMKernel read latency (millisecs) per read.

       GAVG/rd
            The average Guest OS read latency (millisecs) per read.

       QAVG/rd
            The average queue read latency (millisecs) per read.

       DAVG/wr
            The average device write latency (millisecs) per write.

       KAVG/wr
            The  average  ESX  Server  VMKernel  write latency (millisecs) per
            write.

       GAVG/wr
            The average Guest OS write latency (millisecs) per write.

       QAVG/wr
            The average queue write latency (millisecs) per write.

       ABRTS/s
            The number of commands aborted per second.

       RESETS/s
            The number of commands reset per second.

       Interactive Commands

       e    Expand/Rollup storage world statistics.   Allows  viewing  storage
            resource  utilization  statistics broken down by individual worlds
            belonging to a group. You will be prompted to enter the group  id.
            The statistics are per world.

       l    Expand/Rollup  storage  device(=lun)  statistics.  Allows  viewing
            storage resource utilization statistics broken down by  individual
            devices  belonging  to  an expanded world. You will be prompted to
            enter the world id.

       V    Display VM instances only.

       r    Sort by READS/s column.

       w    Sort by WRITES/s column.

       R    Sort by MBREAD/s column.

       T    Sort by MBWRTN/s column.

       N    Sort first by VM column,  then  by  WORLD  column.   This  is  the
            default sort order.

   Network Screen
       This  screen  displays  server-wide network utilization statistics.  On
       this screen statistics are arranged per  port  per  configured  virtual
       network device.  For physical NIC statistics, see the row corresponding
       to the port that the physical NIC is connected to. For virtual NIC con-
       figured  in a particular Virtual Machine statistics, see the row corre-
       sponding to the port that the virtual NIC is connected to.

       Statistics Descriptions

       PORT ID
            The virtual network device port id.

       UPLINK
            Y implies the corresponding port is an uplink.  N  implies  it  is
            not.

       UP   Y implies the corresponding link is up.  N implies it is not.

       SPEED
            The link speed in MegaBits per second.

       FDUPLX
            Y  implies  the corresponding link is operating at full duplex.  N
            implies it is not.

       USED BY
            The virtual network device port user.

       DTYP The virtual network device type.  H  implies  HUB  and  S  implies
            switch.

       DNAME
            The virtual network device name.

       PKTTX/s
            The number of packets transmitted per second.

       PKTRX/s
            The number of packets received per second.

       MbTX/s
            The MegaBits transmitted per second.

       MbRX/s
            The MegaBits received per second.

       %DRPTX
            The percentage of transmit packets dropped.

       %DRPRX
            The percentage of receive packets dropped.

       ACTN/s
            Number  of  actions  per second. This statistic is for internal to
            VMware use only.

       Interactive Commands

       T    Sort by MbTX/s column.

       R    Sort by MbRX/s column.

       t    Sort by PKTTX/s column.

       r    Sort by PKTRX/s column.

       N    Sort by PORT ID column. This is the default sort order.

Batch Mode
       Allows collecting and saving resource utilization statistics in a file.
       Running in this mode is a two step process.
        Step  one  requires  running  esxtop in interactive mode, switching to
       each of the four available  screens,  selecting  the  columns  on  each
       screen  you  are  interested  in  and  saving this configuration in the
       ~/.esxtop310rc file using the W interactive  command.  See  Interactive
       Mode  above.   Note that the global stats are always available in batch
       mode.
        Step two requires running esxtop in batch  mode  and  redirecting  the
       output  to  a  file.  This may be done as follows: esxtop -b > foo.csv.
       Note that the file name needs to have a .csv extension. Esxtop does not
       enforce this, but post-processing tools mentioned next require this.
        Statistics collected in batch mode may be post-processed using such MS
       Windows applications like Excel and Perfmon.
        In this mode, esxtop will not accept interactive commands.  Esxtop  in
       batch  mode  runs  until it produces the number of iterations requested
       (see command-line option n below for  more  details)  or  until  killed
       using CTRL c.
        In this mode there are several command line options available.

   Command-line Options
       b    Run esxtop in Batch mode.

       d    Specifies  the  delay  between statistics snapshots.  Default is 5
            seconds. Minimum is 2 seconds. If a delay of less than  2  seconds
            is specified then the delay is set to 2 seconds.

       n    Number  of iterations.  Collect and save statistics this number of
            times and then exit.

Replay Mode
       Replays resource utilization statistics collected using vm-support (see
       vm-support  man  page  for more information). Running in this mode is a
       two step process.
        Step one requires running vm-support  in  snapshot  mode  on  the  ESX
       Server  Service  Console. This may be done as follows: vm-support -S -d
       duration -i interval.  The resulting zipped tar file needs  to  be  un-
       zipped and un-tar'ed before use by esxtop replay mode.
        Step  two  requires running esxtop in replay mode. This may be done as
       follows: esxtop -R vm-support_dir_path.  Note that it is  not  required
       to run esxtop replay mode on the ESX Server Service Console.
        Replay  mode  can  be run to produce Batch Mode style output (see com-
       mand-line option b below for more information).
        In replay mode, esxtop accepts the same set of interactive commands as
       in  Interactive  Mode.   Esxtop  in replay mode runs until there are no
       more vm-support collected snapshots to be read or esxtop  produces  the
       number  of  iterations  requested  (see command-line option n below for
       more details).
        In this mode there are several command line options available.

   Command-line Options
       R    Specifies the path to the vm-support collected  snapshot's  direc-
            tory.

       b    Run esxtop in Batch mode.

       d    Specifies the delay between screen updates.  Default is 5 seconds.
            Minimum is 2 seconds. If a delay of less than 2 seconds is  speci-
            fied then the delay is set to 2 seconds.

       n    Number of iterations.  Update the display this number of times and
            then exit.

FILES
       ~/.esxtop310rc The personal configuration file.

SEE ALSO
       vm-support(1), cpu(8), mem(8), diskbw(8), numa(8)

DISCLAIMER
       Based on top source code from the OpenBSD distribution.

          Copyright (c) 1997, Jason Downs.  All rights reserved.
          Copyright (c) 1984, 1989, William LeFebvre, Rice University
          Copyright (c) 1989, 1990, 1992, William LeFebvre,
                                          Northwestern University

       Redistribution and use in source and binary forms, with or without mod-
       ification,  are  permitted  provided  that the following conditions are
       met:

       1. Redistributions of source code must retain the above copyright
          notice, this list of conditions and the following disclaimer.

       2. Redistributions in binary form must reproduce the above copyright
          notice, this list of conditions and the following disclaimer in the
          documentation and/or other materials provided with the distribution.

       THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
       IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED  TO,  THE  IMPLIED  WAR-
       RANTIES  OF  MERCHANTABILITY  AND  FITNESS FOR A PARTICULAR PURPOSE ARE
       DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY  DIRECT,
       INDIRECT,  INCIDENTAL,  SPECIAL,  EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES
       (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SER-
       VICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
       CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIA-
       BILITY,  OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
       OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY  OF
       SUCH DAMAGE.




ESX 3.5                          June 7, 2007                        ESXTOP(1)
 
 

サブページ リスト

Comments