Skip to main content

SSH and screen command and JMeter

We all know how useful JMeter distributed testing is when it comes to generating how amount of load from multiple machines. This requires certain ports on remote test agent to be opened so that a centralized JMeter server could talk to them and execute test.
Given the inability to get port opened on aws test agents, I could not employ JMeter distributed testing. Hence I resorted to looping through SSH connection and screen command to execute load test on remote machine. Not an ideal solution but far better than logging in to each test agent and starting test manually.

If you never used screen command then Matt has a quick tutorial on it.
I have load test machine image, it is child’s play to bring up load test agent with exactly same configuration on AWS.

This is how for loop looks like -

for server in ip1 ip2 ip3; do
   ssh -tA <you ssh command here> @$server 'bash -s' < screen.sh
done

screen.sh is shell script which is executed on each ip address mentioned above. you don’t need to copy it to remote server to run it. screen.sh looks like -

# delete existing screen (could be residual from last test)
for session in $(screen -ls | grep -o '[0-9]\{6\}')
do
    screen -S "${session}" -X quit;
done

screen -d -m -S loadtest; screen -S loadtest -X stuff "cd perftest/s2s/;rm testresult*;./runs2s.sh;"`echo -ne '\015'`;

Herein -
<screen -d -m -S loadtest> creates screen with name “loadtest” and detaches from it

<screen -S loadtest -X stuff> is used to pass subsequent command to screen “loadtest” which we created above

<"cd perftest/s2s/;rm testresult*;./runs2s.sh;C-a d" > does following -
  • navigate to required directory,
  • remove the last test results (my test result files are of format testresult_TimeStamp),
  • ./runs2s runs the load test (that is is running JMeter load test from command line) and
  • `echo -ne '\015'` simulates hitting enter key

but why not create session directly using <screen -s loadtest> and send command to it instead of creating a session, detaching from it and then sending commands to detached session ?
I tried it and failed with the error specified here. Solution specified in the post did not work for me and I had to created a detached screen session and send commands to it.

Why use screen.sh and not send command along with ssh as specified here?
I tried this too, but failed on syntax. There are single and double quotes in screen.sh and escaping did not work. If you are able to send it along with ssh command then please post your comment here.

One test execution is over then I use another ssh loop to bring test results to my machine. Which I merge to one file and begin analysis. ssh loop is pretty simple looks as -

for server in ip1 ip2 ip3;
do
   scp -rp $server:perftest/s2s/testresult.* .
done


That’s all :-)

Popular posts from this blog

How to write bad UI automation code (a formal guide)

You might find umpteen resources on learning to write UI automation. But rarely you find any guide on writing bad automation code. Bad automation code is not actually "bad" especially when you consider following - Bad automation code keeps you job safe. Well who can understand the code other than you. Gives you an opportunity to ask for more raise as you are the only one who can maintain the code you write Lets you keep your head up as you the geek of your company who knows that abstruse code  Now coming to writing bad automation code, here are some silver bullets - Record and Replay is the core or writing good (read bad) automation code. It has so many advantages which just out do many others. Biggest being you don't have to write any good code (read bad), tool does it for you. The way to convince decision owners in to enlightening is to let them know that record and replay does not require any special skills. Even your HR can do this.

Using chrome console to test xPath and css selectors

Since the advent of selenium there have been many plugin to test xPath / css selectors but you don’t need any of them if you have chrome browser. Using Chrome console you can test both xPath and css selectors. Launch website to be tested in chrome browser and hit F-12 and you would see chrome console opened in lower pane of application - Hit escape key and console would open another pane to write element locators - And now you can start writing xPath or css selectors in chrome console and test them - The syntax for writing css id - $$(“ ”) And hit the enter key. If your expression is right then html snippet of the application element corresponding to the css selector would be displayed - If you mouse over the html snippet in chrome console then it would highlight the corresponding element in application - If you want to clean console of previously written element selectors then just hit ctrl+L keys and chrome console would be empty again. Pro

Verify email confirmation using Selenium

Email confirmation seems to be integral part of any registration process. I came across an application which lets you provide your email address. You can follow the sign up link in you mail and then complete the registration process. Lets consider we provide GMail address for it. Now if were to use only Selenium then we would have to follow following steps - Launch GMail using Selenium; Some how search for new mail in the list of available mails; Some how click on it; Parse the mail message; Get the registration link; Follow up with registration process What do you think of an approach in which you can