Quantcast
Channel: VMware Communities: Message List - VMware I/O Analyzer
Viewing all 614 articles
Browse latest View live

(Experimental) Scheduling Tests via HTTP Commands

$
0
0

Introduction

In I/O Analyzer 1.5.1, we have introduced an experimental new feature, which allows users to schedule tests via HTTP commands. The idea of such an interface is to provide a mean of script-able test scheduling functionality. In this post, we briefly introduce this new interface and give a few sample use cases.


Note

  • You should still deploy all necessary controller and worker VMs before using this HTTP API.
  • This HTTP API has only minimal error checking, so you should first have all the required information (such as host credentials, worker IP addresses, workload specs, etc) before using this HTTP API.
  • This HTTP API is experimental and is still under development. The API might be changed in future releases.


Communicating with the back-end HTTP script engine
The general command to communicate with the back-end HTTP script engine is as follows:

http://<VA-IP-ADDR>:8180/IOAnalyzer/ioanalyzer/script?<...>


where <VA-IP-ADDR> is the I/O Analyzer controller VM's IP address

and <...> in the end is a subset of the following parameters, concatenated by ampersand '&' (see "supported parameter values" section below).

You may simply open a browser and type in the above URL, or you may use wget in your script/terminal to interact with the back-end script engine.

 

Supported parameter values

  • command=newtest
  • name=<test-name>
  • desc=<optional>
  • duration=<defaul:120>
  • schedule=YYYY-MM-DD_hh-mm-ss
  • readfromfile=[yes|no(default)]
  • host=<hostname>,<password>
  • iometer=<vmIP>,<workloadSpec>
  • replay=<vmIP>,<trace>,<device>,<pace>,[wrap|ignore]
  • savetofile=[yes|no(default)]

All parameter names are in lower case (case sensitive)! Parameters have no order.


Parameter Definitions (all parameters are required unless otherwise specified)

  • command: must be newtest
  • name: test name (same naming rule as web UI)
  • desc: optional test annotation
  • duration: test duration in seconds per workload, 120 if not specified
  • schedule: optional time to run. If not specified, will run at the earliest possible time (i.e., run immediately if no other tests are running).
  • readfromfile: whether to retrieve configs from file (storing on server); default NO; if yes, host/iometer/replay/savetofile parameters will be ignored. (accepts yes, no, 1, 0, true, false; case insensitive.)
  • host: hostname followed by a comma and then the password string. for multiple hosts, repeat this parameter multiple times.
    *Special chars have to be percentage encoded* (e.g., ' '->%20, '#'->%23, '%'->%25, '&'->%26).
  • workload
    • iometer: required for iometer type only. Comma separated VM IP and workload spec.
      For multiple VMs, repeat this parameter multiple times.
    • replay: required for replay type only. Comman separated VM IP, trace name, device name, pace, and wrap or ignore.
      For multiple VMs, repeat this parameter multiple times.
  • Special rule--If neither ioemter nor replay is specified, it will run in listening mode. If both are specified, command fails.
  • savetofile: option to write the config to server. No by default. If yes, it will overwrite the existing config that has the same name. (accepts yes, no, 1, 0, true, false; case insensitive.)


 

Example 1: iometer (please concatenate command in one line)

http://192.168.100.1:8180/IOAnalyzer/ioanalyzer/script?command=newtest&

name=Config-File-1&

savetofile=true&

desc=optional&

duration=300&

schedule=2013-4-20_9-0-00&

host=%3Chostname%3E,`~!@%20%23%25%26$^*()-_=+[{]}\|'%22;:/?.%3E,%3C&

host=192.168.1.1,password&

iometer=192.168.100.1,Max_IOPS.icf&

iometer=192.168.100.2,Max_IOPS.icf

 

Example 1 output

Parsing test: Config-File-1

Desc: optional

Duration: 300 seconds.

Scheduled time: Thu Apr 20 09:00:00 PDT 2013.

Adding host: <hostname>

Adding host: 192.168.1.1

Iometer workload added: 192.168.100.1 Max_IOPS.icf

Iometer workload added: 192.168.100.2 Max_IOPS.icf

Test in iometer mode.

Test saved.

 

 

Example 2: trace replay (please concatenate in one line)

http://192.168.100.1:8180/IOAnalyzer/ioanalyzer/script?command=newtest&

name=Config-File-2&

desc=optional&

duration=300&

schedule=2013-4-20_9-0-00&

host=192.168.1.1,password&

replay=192.168.100.1,DVDSTORE,sdb,1.0,wrap&

replay=192.168.100.2,TPCC-LIKE,sdb,1.0,ignore

 

Example 2 output

Parsing test: Config-File-2

Desc: optional

Duration: 300 seconds.

Scheduled time: Thu Apr 20 09:00:00 PDT 2013.

Adding host: 192.168.1.1

Replay workload added: 192.168.100.1 DVDSTORE sdb 1.0 wrap

Replay workload added: 192.168.100.2 TPCC-LIKE sdb 1.0 ignore

Test in replay mode

 

 

Example 3: listening mode

http://192.168.100.1:8180/IOAnalyzer/ioanalyzer/script?command=newtest&name=Config-File-2&duration=300&&host=192.168.1.1,password

 

Example 3 output

Parsing test: Config-File-2

Duration: 300 seconds.

Scheduled time: Thu Apr 20 12:56:58 PDT 2013.

Adding host: 192.168.1.1

Test in listening mode.


Report bugs

Please report bugs to io-analyzer-info (at) vmware.com.


Unable to add workload entry

$
0
0

Curious to know if you have any input on a problem in which we can successfully add the ESX hosts while in the “master” analyzer VM and the I/O Analyzer VMs are correctly listed from the ESX hosts added, but when we select the worker VMs from the drop-down in the add workload entry section nothing happens after selecting the VM and then clicking “Add”.  Nothing happens after clicking the “Add” button (good, bad or otherwise).

Re: Unable to add workload entry

$
0
0

Just had a few mails with Mark and it looks like the problem is because of cloning IOAnalyzer VMs. Mark, please let us know if you think there's some improvement we can make on this issue.

 

Thanks,

Chien-Chia

Re: Unable to add workload entry

$
0
0

Thanks again for the quick reply on this one.  Not sure I would spend much time trying to figure out why the clone operation seemed to possibly cause the problem as it was very easy to just deploy from the ova multiple times.  If you have extra time it may be good to validate what we saw and add a blurb about it being best to deploy fresh from the ova in place of cloning.  Thanks!

Problem: Trace characteristics display no charts

$
0
0

I had uploaded a remote trace file, the io analyzer notified me a successful result, but Trace characteristics display no charts:

Workload Name : exchange_loaddfdf

Trace duration is 0 Seconds

Trace contains 0 requests

Logical Block Numbers accessed in the trace ranges from to

Minimum disk size required to replay trace accurately: 0 GB

Trace contains 0% reads and 100% writes.


But, using vscsiStats -e trace_file command, I can get the trace contents:

write,2147483713,4096,1,215738104,627935733713,8207

write,2147483662,393216,96,23931272,627936635248,16392

read,2147483675,49152,12,227885808,627937646320,12178

write,2147483723,4096,1,224847424,627940438783,13619

 

Can anyone tell me what's the problem? Thanks.

What do the fields in a trace file mean?

$
0
0

The vscsiStats help says:

"

- trace contains NO customer sensitive data

- only information recorded is:

                  - serialnumber, IO block size, number of scatter-gather elements

                  - command type, block number, timestamp"

"

 

Take this trace information "write,2147483713,4096,1,215738104,627935733713,8207" as a example,

write               of  course this is command type

2147483713    serialnumber? what does it mean?

4096              IO block size?

1                   block number?

215738104    ?

627935733713    timestamp?

8207             ?

 

Is there anyone could help me, thanks.

Re: How i get rid of of old test results

$
0
0

By IO analyzer 1.5.1, after the test result was archived, I can't find the archived report under /var/www/archives directory, but it was still under /var/www/expts directory, but using "." prior to original dir name to hide it.

 

Another problem, after archive, I can never review the test report from the web page, is there any suggestion? thanks.

NFS Device Statistics

$
0
0

I'm currently working on validating a new vSphere 5.1 environment predicated on 10Gbit NFS.

 

When I'm using resxtop I can see the device statistics for the NFS volumes (CMD/s, READS/s,WRITES/s, etc.....GAVG/cmd - obviously not DAVG, KAVG or QAVG), but it appears that IO analyzer doesn't capture these, only showing me the device statistics for the local storage in each ESXi server.

 

Is there anyway of modifying it to display this data, or can I submit a feature request for a future release?

 

Thanks,

 

Ben


Re: How i get rid of of old test results

$
0
0

When displaying results, the backend will skip all hidden folders. If you would like to restore the archived results, please rename the folder to remove the leading dot ('.').  If you would like to delete the results persistently, please delete the hidden folders.

 

Thanks,

Chien-Chia

Re: NFS Device Statistics

$
0
0

Our backend doesn't request NFS client stats so for the current version, there will be no NFS stats. We'll add this feature in the next release.

 

Thanks,

Chien-Chia

Re: Problem: Trace characteristics display no charts

$
0
0

I just tried it and was unable to reproduce your problem. One thing I notice though is on p.20 of the IOAnalyzer User's Guide (Collecting a vSCSI Trace), the command in step#7 misses an output redirection symbol ('>'). The correct command should be

 

vscsiStats -e <temporary-file-name> ><trace-file-name.csv>

 

After the command, please use text editor to examine the CSV output file content, which should start from the following line:

#Vscsi Cmd Trace. (Trace Format Version 2)

 

followed by the definitions of each column:

#Command Type,Serial Number,IO Data Length,Num SG Entries,LBN,Time Stamp (microseconds),Command Latency (microseconds)

 

The 4 line snippets you posted seem to be in the correct format. So please simply make sure the output redirection symbol is added to your "vscsiStats -e" command and it should work.

 

Thanks,

Chien-Chia

Re: What do the fields in a trace file mean?

$
0
0

The definitions of each column can be found in the second line of your CSV output file:

#Command Type,Serial Number,IO Data Length,Num SG Entries,LBN,Time Stamp (microseconds),Command Latency (microseconds)

 

Thanks,

Chien-Chia

Re: How i get rid of of old test results

Re: Problem: Trace characteristics display no charts

$
0
0

Dear Chien-Chia, you'r right , thank you very much.

 

I didn't read the user's guide, but only refer to the vscsiStats help:

- If successfully started, log channel id(s) will be printed out.

                To store the command trace in a file for later processing, invoke:

                $ logchannellogger <log_channel_id> <binary_trace_file>

 

without converting the <binary_trace_file> into csv file, resulting to the mentioned problem.

Re: What do the fields in a trace file mean?

$
0
0

Dear Chien-Chia,

Your're right, thanks again.


Re: NFS Device Statistics

$
0
0

That's great - thank you very much.

I/O Analyzer 1.5 - Tests Not Running

$
0
0

I just deployed the I/O Analyzer 1.5 appliance on a vSphere 4.0 host (ESX, not ESXi) version 4.0.0 build 208167 and created a configuration to test Max Throughput and had it run immediately.  I tried this a couple times, rebooting the appliance in between attempts and the tests will not complete.  They begin running with a status of "Running (initializing VI SDK)" and then just hang there and never error out or complete.  Has anyone seen this behavior before?  Any recommendations on how to fix the problem?

 

I have used the appliance before on a newer vSphere 5.0 test system without issues, this is the first time I have attempted to run it on 4.0, also the previous tests were run using version 1.1 of the appliance.

 

Thanks,

Brian

Re: I/O Analyzer 1.5 - Tests Not Running

$
0
0

Hi Brian,

 

Could you please send /var/www/ioa.log and /var/log/daemon.log to io-analyzer-info@vmware.com? It looks like this is a problem related to VI SDK.

 

Thanks,

Chien-Chia

Re: Add first host

Re: Add first host

$
0
0

This sounds like the I/O Analyzer VM was unable to retrieve host information via VI SDK. Could you please double check from a terminal in VI client consolse, you are able to ping ESX?

 

Thanks,

Chien-Chia

Viewing all 614 articles
Browse latest View live


<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>