How to solve this issue in VMmark test?

[williamtw]

When we run Vmmark Harness, It always failed and show the message “Could not PING the following 6 machines”.

(Please see attached file harness_run_fail.JPG)

Although I confirmed that the Client could ping 6 machines, the test still got the same failure.

Does anybody know how to solve this issue? Thanks!

[lroderic]

Hi. There's IP ping, and then there's STAF ping. STAF uses its own ping command to check that all hosts in the testbed are ready to send/receive STAF commands. The CLI equivalent is:

STAF machinename ping ping

The error you received indicates that STAF is not installed or configured correctly. See the section titled "Install and Configure the STAF Package" in the guide for instructions on configuring STAF on the clients and all VMs in the testbed.

If security isn't an issue, you can use the following shortcut in STAF.cfg rather than explictly "trust" each machine as in the sample STAF.cfg we provide:

trust level 5 machine tcp://10.1.19.

Don't forget to download the javaserver_function.xml patch file at . You'll need it.

Lisa

[williamtw]

Thank you, Lisa.

But I don't quite understand what you mean.

Could you please tell me in detail?

Also, could you please help me check my setting in the attached file if possible?

PS. we run one-tile benchmarks.

Thanks!

[simonpha]

I had this exact same issue at least with respect to the Windows VMs.

STAF doesn't start correctly because the STAF.cfg file for windows needs to have some additional lines commented out for all but the primeclient. On your STAF.cfg_windows file, the last section shows

-

---

service stax library JSTAF execute \

C:\STAF\services\stax\STAX.jar \

parms "numthreads 100 logtcelapsedtime enabled logtcstartstop enabled"

service event library JSTAF execute \

C:\STAF\services\stax\STAFEvent.jar

-

---

Comment out those additional lines as it says in the comments above it and then STAF will start on those Windows systems with no errors.

You would have to look further into why your linux VMs are not starting the STAF process. But that is why they cannot ping each other through STAF.

[williamtw]

Thank you Sir,

We followed your suggestion to modify the STAF.cfg files in every virtual machine but still got the same result.

Do you have any other suggestions?

[jamesz08]

Is the STAFProc process running in all your VMs?

If you do not see the STAFProc process you need to start it.

If you see the STAFProc process running should be able to run the following command from the client cmd window:

staf standby0 ping ping

What is the response to that?

[lroderic]

Hi. As James said, make sure STAFProc is running. Go into Windows Task Manager on the prime client and look for a STAFProc.exe process. If you don't see one, open an MS-DOS command window and start STAF:

c:\STAF\bin\STAFProc.exe

Then minimize this window. Don't close this window since that will kill STAFProc.exe.

Lisa

[williamtw]

Thanks James,

I tried the command "staf CLIENT ping ping"

The following clients was responded "PONG"

standby0

mailserver0

javaserver0

fileserver0

But database0 and webserver0 was responded as following

Error submitting request, RC: 16

Additional info

-

---

STAFConnectionProviderConnect: Timed out connecting to endpoint: select<> timeout: 22, Endpoint: tcp://database0

I checked these files hosts, staf.conf and boot.local, but I couldn't find out any problems.

What should I do? Thank you!

[jamesz08]

Your database and webserver VMs do not have STAF running.

Run the command: /etc/init.d/start-staf.sh

That will start the STAF process. You will then be able to run the ping command successfully. This command should be included in the /etc/init.d/rc.local file which is executed when the VM boots so it starts automatically.

Once all the VMs pass the ping you can try running the VMmark harness again.

[williamtw]

Dear James,

We installed STAF 3.34 and failed to start up the STAF.

Input this command ./etc/init.d/start-staf.sh and check this file

/tmp/start-staf.out, it will show this message below.

Error on INTERFACE definition, interface tcp library STAFTCP

Error creating interface, RC: 47, Reason: STAFConnectionProviderConstruct:

Could not determine logical/physical identifier.Error code: 22 Reason:

Error getting hostent structure for host name: fileserver, gethostbyname()

RC=2

Error reading config file

If we comment out this line "interface tcp library STAFTCP option" in

STAF.cfg, we were able to start up the STAF.

Check this file /tmp/start-staf.out, it will show this message below.

Machine : local

Machine nickname : database0

Startup time : 20100202-09:54:10

STAFProc version 3.3.4.1 initialized

However, even though the STAF was started up by this way, we still caludn't

ping out to the client via STAF.

Did you ever see this issue? What should I do?

Thank you!

[jamesz08 · Accepted solution]

Are you using the template VMs or did you build your own? The template VMs should have staf already installed and working.

You cannot ping because the TCP library is not loading. You need to be able to start STAF with that TCP library. It looks like it cannot resolve the hostname fileserver, make sure the hostname is set in your /etc/hosts file with the correct IP for your configuration.

[williamtw]

Dear James,

I was able to ping STAF successfully after using the template VMs.

But I still could not complete this test. It will show "MailServerTile0 fail setup" (Please see attached file mailserver_filed.jpg)

Did you ever see this issue? What should I do?

Thank you!

[mitchc]

Hi!

First off, make sure you followed the steps to run LoadSim and populate the MailServer data, and it ran without errors. It is sometimes necessary to bring up your mailserver0 and wipe out the mail databases if they somehow became corrupted, then re-run LoadSim once you've setup lpriv1 and lpriv2 fresh. Play around with making sure LoadSim is completely happy before you run staxmon.jar again. It should be able to run against the vmmarktemplate.sim file in c:\vclient0\mailserver without errors.

The other thing to mention, is that STAF is a tricky beast.. so make sure it is set up to start automatically in your windows boxes on behalf of the domain administrator.. possibly something like "MAILSERVER00\Administrator". Note this needs to be followed on mailserver0 and client0, as even if you follow the instructions in the guide carefully you might end up starting STAF automatically as "MAILSERVER0\Administrator" (note only one zero, not two!) and "CLIENT0\Administrator", which are both probably wrong! I'd recommend process explorer for investigating what processes are running on a windows box.

cheers!

Mitch.

[williamtw]

This issue has been solved.

Thank you!