evaperf有兩種用法,一種是和windows下的performance monitor工具結合,加了一些計數器,使用時可以使用perfmon開啟performance monitor工具,然後點擊工具列裡邊的 + 號,加入你喜歡看的東西,如Total Host req/s, Total Host Kb/s等等,適時的監控EVA效能。
但這一種方式有個麻煩的地方就是不能拿一個長時間的report。而且同時能觀察的類型有少了些。
第二種是命令列的方式,這一種方式可以收集長時間的資料,而且可以收集到所有EVA各個方面的資料,但缺點是您需要把收集到的資料拿回來,自己做處理(您可以採用多種辦法來處理這些資料,比如unix下的awk, grep,還有一個好象叫 tlviz, 俺沒有用過,因為俺更願意在unix下過濾出自己想要的東西,然後在excel下畫一些表格,看一下效能狀況。下面的步驟是採用 evaperf來收集效能資料的步驟。
1。進入dos 提示符下,cd到以下目錄:
C:/Program Files/Hewlett-Packard/EVA Performance Monitor>
evaperf命令列的所有東西都在這個目錄下。這個目錄下有一個可執行程式evaperf, 就是我們主要要使用的工具。直接敲evaperf, 不加任何參數,產生協助資訊如下:
C:/Program Files/Hewlett-Packard/EVA Performance Monitor>evaperf
Usage:
ls List storage arrays
as Array status
cs Array controller status
pd Physical disks
pdg Physical disk groups
pda Physical disk activity map
vd Virtual disks
vdg Virtual disk groups
vdrl Virtual disk read latency histograms
vdrlg [lunwwn] Virtual disk read latency histogram graph
vdwl Virtual disk write latency histograms
vdwlg [lunwwn] Virtual disk write latency histogram graph
vdts Virtual disk transfer size histogram graph
vdtsg [lunwwn] Virtual disk transfer size histogram graph
hc Host connections
ps Port status
hps Host port statistics
drt Data replication tunnels
luns LUNS visible to this host
rc wwn Reset the counters for an array
fnh host username password
Add a new CV/EVA host to the list
fnh Show the list of known CV/EVA hosts
fn Reload friendly-names from known CV/EVA hosts
sfn Show the friendly name map
spw wwn password Set the access password for an array
dpw wwn Delete the access password for an array
vpw Verify array passwords
all Runs: ls, as, cs, vd, vdg, hc, ps, hps, pd, pdg, drt
Windows Performance Monitor filtering commands:
pfs Show the Perfmon filter configuration
pfd Delete the Perfmon filter configuration
pfa [array]* Set the Perfmon array filter list
pfvd [vdisk]* Set the Perfmon vdisk filter list
pfh Help for the Perfmon filter commands
Filtering and formatting options:
-sz array [array]* Only collect data from the arrays specified in the list.
You must supply as least one array name.
The array name may be either its WWN or its friendly-name.
-fvd vdisk [vdisk]* Only collect data from the virtual disks with the names specified in the list.
You must supply as least one virtual disk name.
The virtual disk name may be either its WWN or its friendly-name.
This affects the vd, vdrl, vdwl and vdts commands only.
-cont [N] Continuous at N second intervals. Default is one second
-dur N Continuous mode duration in seconds
-csv Output in comma-separated variable format with timestamp
-tsv Output in tab-separated variable format with timestamp
-nh No table headings
-ts1 Add time stamp to CSV/TSV like: Fri Jul 23 16:23:05 2004
-ts2 Add time stamp to CSV/TSV like: 23/Jul/2004 16:23:05 (default)
-nots No timestamp in CSV/TSV output
-fd keyword [keyword]* Output data lines that contain ALL keywords
Note that data is still collected and processed for all arrays.
This just filters the table ouput rows.
-cn Use name contractions in fnames.dict
-KB Show rates in KB/s (default is MB/s)
-us Show times in microseconds (default is milliseconds)
-nfn No friendly names (show node or LUN WWN)
-fo filename Write output to filename as well as console
Version: 1.00.50729 (Built Fri Jul 29 10:58:05 2005)
Please note:
MB/s values are based on 1MB = 1,000,000 bytes.
KB/s values are based on 1KB = 1,024 bytes.
Where a command accepts an array or LUN WWN, you may enter a friendly name
instead of the WWN. Friendly names are case sensitive. WWNs may be entered with
or without hyphens.
2。運行以下命令
C:/Program Files/Hewlett-Packard/EVA Performance Monitor>evaperf fnh xx.yy.zz.dd administrator administrator
Verifying access ...
Host Username Accessible
----------- ------------- ----------
xx.yy.zz.dd administrator Yes
這個命令的目的串連到sma, 準備產生friendly name, 因為沒有friendly name時,您在產生的report上看到的不是vdisk01,vdisk02等,而是冗長的wwid, 看起來費盡。
其中:
fnh是evaperf 產生friendly name的參數。
xx.yy.zz.dd 是你的sma小家電IP地址
administrator 和hpinvent是command view eva的使用者名稱和密碼,如果管理小家電是storage server 2003, 密碼是hpinvent. 如果是sma, 密碼是administrator。
3。運行命令產生friendly name。
C:/Program Files/Hewlett-Packard/EVA Performance Monitor>evaperf fn
Copying fnames.conf to fnames_conf.bak
Attempting to load names from host: xx.yy.zz.dd
ARRAY 5000-1FE1-5000-FA21 szeva1
Found 2 virtual disks
VDISK 6005-08B4-0010-2170-0002-5000-005D-0000 VDtest1
VDISK 6005-08B4-0010-2170-0002-5000-0069-0000 test_dsk1
Fetching host data for 5000-1FE1-5000-FA21 eva1
Found 2 hosts
HOST 1000-0000-C93B-B99B fs25
HOST 1000-0000-C931-15B1 aixp690
Fetching disk group data for 5000-1FE1-5000-FA70 eva1
Found 1 disk groups
DISKGROUP 5000-1FE1-5000-FA21:0 test1
4。使用以下命令列出可以看到的eva名字
C:/Program Files/Hewlett-Packard/EVA Performance Monitor>evaperf ls
The HP EVA Performance Data Collection Service (evapdcs) does not seem to be run
ning
Attempting to start the service
The service has been started
Device Path Target LUN Product Product Ctlr Serial Harware
Name Node
ID ID ID Rev. Ver.
---------- ---- ------ --- ----------------
------------------- -------------------
//./Scsi3: 0 0 0 HSV110 (C)COMPAQ 4001 S045 P584901AAQS040 110
5000-1FE1-5000-FA70 5000-1FE1-5000-FA70
//./Scsi3: 0 1 0 HSV110 (C)COMPAQ 4001 I07O P5849D3AAOI07O 110
5000-1FE1-5000-FA70 5000-1FE1-5000-FA70
5。開始收集效能資料。
剛才的協助檔案部分大家也看了,有不少參數,很容易迷糊,但希望不要被現象所蒙蔽,個人認為大部分參數你可能不會關心,因為一般情況下,我們希望收集到所有的效能資料。我把我常用的命令列出來解釋一下,大家就很清楚怎樣使用後邊的各種參數的。
C:/Program Files/Hewlett-Packard/EVA Performance Monitor>evaperf all -ts1 -tsv -sz eva1 -cont 20 -dur 86400 -fo guo.2
解釋:
all: 意思是我要收集所有的資料,包括vdisk, controller,phicial disk, disk group 等。
-ts1: 是我給輸出報告加時間標籤的格式為Fri Jul 23 16:23:05 2004
-tsv: 是需要加上時間標籤,沒有這個選項的話,輸出紀錄無時間標籤。
-sz eva1: sz 不是深圳,是選擇我要收集資料的eva的名字是eva1, 因為大家都知道,一台小家電可能管理多台eva,每台名字不一樣。
-cont 20: 是每隔20秒收集一次,這個值不要太小,具體原因好像白皮書上有提到過。
-dur 86400: 一共收集24小時,也就是86400秒
-fo guo.2: 輸出檔案到guo.2,同時也輸出到您的dos螢幕上。
6。分析
收集完之後,取出guo.2文字檔,然後自己分析,分析的工具之前也提到了,大家可以根據自己的愛好來選用。俺是在unix下邊做的處理。