JMeter CSV Data Set Config: File not found & Sharing Mode

CSV Data Set Config: File not found & Sharing Issues

Symptom

When executing a test in CLI mode, CI/CD pipelines, or distributed testing, one of the following errors occurs in jmeter.log:

2026-08-18 14:02:11,551 ERROR o.a.j.c.CSVDataSet: java.io.FileNotFoundException: users.csv (The system cannot find the file specified)
  at java.io.FileInputStream.open0(Native Method)

Or a test data logic error occurs:

• All concurrent virtual user threads read the exact same row of data simultaneously.
• The test terminates unexpectedly early because of an unexpected End of File (EOF).
• Dynamic CSV variables appear literally as \${username} in HTTP requests.

Quick diagnosis (TL;DR)

• FileNotFoundException in CLI/CI: JMeter resolves relative CSV paths relative to the current working directory from which the jmeter command was executed, not necessarily the directory containing the .jmx file.
• Duplicate Data Read: The CSV Data Set Config is placed inside a loop or child controller, resetting file pointers on each iteration, or Sharing Mode is misconfigured.

Common causes

Issue	Cause	Fix
CLI FileNotFoundException	Invoked jmeter from a folder other than the test plan folder	Use \${__P(testdata.path)} or FileServer basedir
Distributed testing FileNotFoundException	CSV file exists on controller machine but was not copied to remote worker nodes	Copy the CSV file to the exact same file path on all remote worker machines
All threads reading same line	CSV Data Set Config placed as child of a Loop Controller	Move CSV Data Set Config to the root of the Thread Group
Test stops prematurely	”Stop thread on EOF” set to true on a small CSV file	Set “Recycle on EOF” to true and “Stop thread on EOF” to false
Variables not assigned	”Variable Names” blank and CSV has no header row	Provide comma-separated variable names in configuration

Fix (ordered)

1. Use Robust Relative Path Resolution

To make .jmx test plans fully portable across local GUI, CLI, and CI/CD runners regardless of current working directory:

In CSV Data Set Config -> Filename, use the FileServer basedir function:

${__groovy(org.apache.jmeter.services.FileServer.getFileServer().getBaseDir())}/users.csv

Or configure a dynamic property with a default fallback:

${__P(data.dir,${__groovy(org.apache.jmeter.services.FileServer.getFileServer().getBaseDir())})}/users.csv

Now you can override data paths dynamically in CI/CD:

jmeter -n -t /plans/checkout.jmx -Jdata.dir=/ci-data/ -l results.jtl

2. Configure Sharing Mode for Concurrent Threads

The Sharing Mode setting controls how file pointers advance across virtual users:

Sharing Mode	Behavior	Best for
All threads (Default)	Single global file pointer shared across all thread groups. Each thread reads the next unique line	Unique user logins, single-use promo codes
Current thread group	Each Thread Group maintains its own independent file pointer	Separate scenarios that need the same test data set
Current thread	Each individual virtual user reads the entire CSV file from line 1 independently	Iterating through a fixed list of product IDs per user

3. Handle End of File (EOF) Behavior

In CSV Data Set Config:

• Recycle on EOF?: Set to True if you want threads to loop back to line 1 when the file ends.
• Stop thread on EOF?: Set to False to allow tests to run for their scheduled duration. Set to True only if the test must stop immediately after exhausting the data file.

4. Fix CSV Distribution in Distributed / Remote Testing

In distributed mode (jmeter -r):

1. JMeter does NOT send CSV files over RMI to worker nodes.
2. You must place the CSV file on every remote worker machine at the exact same path.
3. If using shared network storage (NFS/EFS), ensure all worker pods mount the share at the identical mount path.

Related tools and topics

Resource	Use when
GUI works, CLI fails	File path and working directory differences
Functions and variables	Built-in functions like __CSVRead
Distributed testing	Managing test data across remote workers
CLI Command Builder	Generate CLI commands with dynamic data properties

Frequently asked questions

Should I use CSV Data Set Config or __CSVRead() function?

Always prefer CSV Data Set Config. It is significantly faster, thread-safe, supports multi-threaded sharing modes, and handles quotes and delimiter parsing properly.

How do I split one CSV file across multiple worker nodes?

Split the CSV file into chunks before the test run (e.g. split -l 1000 users.csv worker_) and copy chunk 1 to Worker 1, chunk 2 to Worker 2.

Continue Learning

→

Next Practical Step

Use \${__groovy(org.apache.jmeter.services.FileServer.getFileServer().getBaseDir())}/file.csv in the CSV Filename field for bulletproof path resolution.

📖

Related Reference

• Functions and variables
• Distributed testing
• Best practices