8 days ago
Topic:
Potential Bug with Server Authentication
George @StresStimulusAdministrator Posts: 397
|
It looks like you reported an issue that your script works on lower load levels and fails on higher load levels. Further, you suggest that a potential StresStimulus bug causes the issue because your other script works on the higher load levels.
Generally, tests can fail for the following two main reasons:
Script misconfigurations Application, instability, or failure to handle the load.
I didn't find anything in your post that would lead to the suspicion that StresStimulus has a bug.
I recommend contacting your support rep and seek assistance with your script configuration. However, if you still believe that you found a bug, please provide a methodology for replicating the issue.
- Cheers
|
|
|
8 days ago
Topic:
Potential Bug with Server Authentication
mwattePosts: 10
|
There seems to be a potential bug running a load test of 3 different applications, on the same host, simultaneously following the below scenario:
Test Case 1 - Application 1 utilizes form authentication
Test Case 2 - Application 2 utilizes form authentication
Test Case 3 - Application 3 utilizes server authentication
When running a Steady Load of 30 users - 10 per application, all users are authenticated successfully.
When increasing the load test to 45 users - 15 per application, all users utilizing form authentication are authenticated successfully. However, the test case utilizing server authentication, only a subset of users are getting authenticated, meaning not all 15 users are authenticated.
Just to note, the application using server authentication (Test Case 3) can handle more than 15 users as I was able to successfully run a load test of 50 users just executing this test alone.
Version 5.3.7670 x64 CLR 4.0
|
|
|
9 days ago
Topic:
Adding Load generator error
George @StresStimulusAdministrator Posts: 397
|
The controller of the machine is having connection issues to the load agent machine. It is most likely caused because of firewall on the agent machine. Make sure the port 49998 is open on the firewall on the agent machine.
|
|
|
9 days ago
Topic:
Adding Load generator error
Unregistered UserPosts: 393
|
When I am trying to add one Load generator receiving below error –
Checked both system ports. Same error. Pls help me to remove the error.
|
|
|
12 days ago
Topic:
How to avoid duplicates when using dataset?
George @StresStimulusAdministrator Posts: 397
|
This behavior is by design and here is why and how to avoid it:
I am assuming you are using VU-Iteration-Bound binding for your dataset. The formula that calculates which record will be used (https://support.stresstimulus.com/display/doc53/Databinding ) is
Record Row number = (Iteration #) * (Max VUs) + (VU #)
When you set the How to count iterations property to Total, you do not guarantee that every iteration will get an equal number of iterations. If VU1 executes ten iterations, then on the 10th iteration, it will use the 301-st record, but since you only have 210 records, the scripts will re-use the top records.
I suggest setting the VU-to-row mapping to On-Demand described https://support.stresstimulus.com/display/doc53/Default+databinding and try rerunning the test.`
|
|
|
13 days ago
Topic:
How to get transaction response times for each act
George @StresStimulusAdministrator Posts: 397
|
There are two ways to get the transaction response times for each iteration:
• You can query the database and open the Activity table that is described https://support.stresstimulus.com/display/doc53/Activity
• Open the Activity diagram for a particular transaction -> right-click > select Export to CSV… to save the .csv file with all iteration times for the transaction https://support.stresstimulus.com/display/doc53/VU+Activity+View
|
|
|
16 days ago
Topic:
How to get transaction response times for each act
Unregistered UserPosts: 393
|
Please let us know if we can get a raw results report from StresStimulus. We are looking for transaction response times for each action in each iteration.
|
|
|
21 days ago
Topic:
How to avoid duplicates when using dataset?
Unregistered UserPosts: 393
|
I was getting in touch as our team may have found a potential bug. Can SS confirm whether they are seeing the same issue please? Below are the details:
Details:
- Vusers: 30
- Dataset entries: 210
- Limit Dataset: Yes and “0”, i.e. all of the data was available to each VU.
- In our recent performance tests we have set “How to count iterations” to “Per VU”. This has shown the VUs, picking unique data from the dataset with each iteration.
- However, we recently changed the setting to the below. In theory, we assumed by choosing this option we wouldn’t need to calculate the number of “Max Iterations” per VU. Because we would simply enter the total number of iterations the all vusers should complete.
- What we discovered when selecting this option is the first 15 VUs would select unique rows data from the dataset, however, after that point the next VU seemed to pick the first row from the dataset. This caused duplicate orders to be created with the same subject details.
- By the end of test this setting achieved the goal of completed the Max Iterations set. However, during the process it ended up creating duplicate orders.
Can SS confirm whether our assumption on how this option works is correct?
|
|
|
3/28/2021
Topic:
How to avoid login bottleneck.
George @StresStimulusAdministrator Posts: 397
|
The reason for the “ļogin storm” is because you probably use the steady load pattern where all virtual users are instantiated simultaneously https://support.stresstimulus.com/display/doc52/Steady+Load
To avoid the login bottlenecks, you can instantiate virtual users sequentially. To do that, use the step load pattern https://support.stresstimulus.com/display/doc52/Step+Load
For example, you can use the following settings:
Start VU 1
Sep VU increase 1
step duration (s) 20 20
Max VU N
This way, the login transaction for VUs will space out by 20 seconds.
|
|
|
3/28/2021
Topic:
How to avoid login bottleneck.
Unregistered UserPosts: 393
|
We were able to create testing scenarios, however we have some problems during initial logins from some computers used for test. Is it possible to delay login procedure not to cause “ļogin storm” from testing computer?
|
|
|
3/26/2021
Topic:
Using different credentials on different iteration
George @StresStimulusAdministrator Posts: 397
|
It sounds like now you need to change a single VU’s server authentication credentials on different iteration.
In server authentication, the credentials are stored on the connections level, so you would need to close the connection and reopen it with new credentials on every next iteration.
It can be accomplished by using a Reset Connection object (described http://support.stresstimulus.com/display/doc53/Reset+connections)
First, you would need to add the Reset-connections object to the very end of your test case.
Then in your Authentication section set the VU-to-Row mapping = On-Demand
|
|
|
3/16/2021
Topic:
Running correlation finder got stuck on 1 request
George @StresStimulusAdministrator Posts: 397
|
StresStimulus autocorrelation is built on a proprietary algorithm that entails scanning your script and evaluating multiple regular expression patterns. In some rare cases (for example, when an HTTP message is bloated or has irregular content like binary), scanning a pattern for a particular request may take an abnormally long time.
When you click "Skip," StresStimulus will skip all remaining patterns for this request. However, it will still try to complete the currently running pattern.
I suggest giving it up to 5-10 minutes, if you can, for autocorrelation to complete. If it still doesn't complete, then you can exclude this session from autocorrelation altogether. Here are the steps:
In your case, it's session #24, which has a path /s159N-Dashboard/. Go to main menu > Edit > Settings > Autocorrelation tab Type /s159N-Dashboard/ into Do not create extractors from these pages textbox (see below):
Click OK and re-run autocorrelation or rerecord the test case.
|
|
|
3/16/2021
Topic:
Running correlation finder got stuck on 1 request
Unregistered UserPosts: 393
|
For one of my script, the auto-correlation is stuck always at one point (Request) of the script. It doesn’t move forward even if I skip for that request.
I have not come across this issue while recording the scripts for other flows multiple times.
I am not sure what is causing this halt on auto-correlation. Can you please throw some light here?
Screenshot below:
|
|
|
3/11/2021
Topic:
How to avoid root certificate installation?
George @StresStimulusAdministrator Posts: 397
|
The load testing process can be broken down into two phases:
• Phase I: recording a script;
• Phase II: executing all subsequent steps such as verifying and configuring the script, running tests, and analyzing results.
Phase I - requires a Fiddler root certificate (or a similar StresStimulus root certificate) for recording. However, once the recording is complete, Phase II does not need the certificate.
So here is a a possible workaround for you:
1. Record your script from a different environment where you can install the certificate.
2. Transfer your script into your restricted environment.
3. Execute phase II in the restricted environment without the certificate.
|
|
|
3/11/2021
Topic:
Graph Sample Rate
George @StresStimulusAdministrator Posts: 397
|
The graph section displays curves consisting of Data points that represent instant performance snapshots collected at specific intervals.
The graph sample rate defines the frequency of the snapshot. Once the test is over, there is no way to change the sample rate as data points are already collected. Because the intervals of collecting data points are not synchronized with the beginning or end of transactions, determining the average performance characteristics of a transaction by datapoints is not the best approach.
You can use instead of the Transaction’s Activity diagram and exporting all transaction times into .csv, as described here.
This information will present performance characteristics for every user executing the transaction and iteration.
- Cheers
|
|
|
3/11/2021
Topic:
How to avoid root certificate installation?
Unregistered UserPosts: 393
|
Environment where we want to run tool has a lot of restrictions and doesn't allow us to install root certificate for Fiddler.
Is there a way for us to run load test for it's endpoint without installing Fiddler?
|
|
|
3/8/2021
Topic:
Graph Sample Rate
sachpatelPosts: 9
|
Hi,
After a test has completed, is there a way of editing the graph sample rate within SS? I am currently exporting the Transactions graph data csv file into Excel and filtering the rows on a multiple of a number, i.e. 20 seconds. Is there an easier method?
Thank you.
|
|
|
2/26/2021
Topic:
Concurrency mode
George @StresStimulusAdministrator Posts: 397
|
Since you contacted support they will reply directly.
- Cheers
|
|
|
2/25/2021
Topic:
Concurrency mode
sachpatelPosts: 9
|
sachpatel wrote:
Does SS have a way of dynamically handling the way requests are processed different with each iteration?
Different in what way? Can you please clarify the question. StresStimulus handles requests according to the recorded script, your script settings and your real-time server responsiveness.
I was referring to how the waterfall model can differ with each recording.
I have sent an email to the support team after running several tests with different settings, as recommended, and provided the outcome. If you could please refer to that and provide any further suggestions of options to try, that would be much appreciated. Thank you.
|
|
|
2/25/2021
Topic:
Concurrency mode
George @StresStimulusAdministrator Posts: 397
|
sachpatel wrote:
Is there a best practice in terms of recording a script to capture accurate concurrency for the application?
StresStimulus recorder is designed to automatically records correct concurrency.
sachpatel wrote:
Because when I do different recordings of the same steps, the waterfall diagrams are not necessarily the same.
You are correct, two recorded waterfalls won’t be exactly the same due to multiple random factors. So the response time will not be exactly the same. However, the waterfall should have sufficient information for a rough comparison of the browser and StresStimulus concurrencies.
sachpatel wrote:
Does SS have a way of dynamically handling the way requests are processed different with each iteration?
Different in what way? Can you please clarify the question. StresStimulus handles requests according to the recorded script, your script settings and your real-time server responsiveness.
sachpatel wrote:
We recently faced an issue where scripts were ported from another test environment and a different version of CRM. In that scenario it would be worth checking the waterfall diagrams before re-scripting?
The main reason for checking the waterfall diagrams is if you want to double-check that your script settings are correct. Once your script is configured, you can run your test multiple times without changing settings or checking the waterfall diagram.
- Cheers
|
|
|