2.4.3

• Fix crash on windows (#1924)

2.4.2

• Add --expect-workers-max-wait parameter (#1922)
• Track worker memory usage (#1917)
• Other small fixes

2.4.1

Fix stat printing when using shapes #1907

2.4.0

Add start_time and url parameters to request event. #1900
Support (and test) Python 3.10 #1901
Make User.run/TaskSet.run final and raise an exception if someone marks it with @task #1895
Ensure target_user_count is set before ramping-up or down #1891/#1894
Release docker image for arm64. #1889

2.2.3

Fix issue with custom arguments in config file (when not running headless) #1888

2.2.2

• Fix version in Docker builds
• Automated change log generation is broken. Will fix this later, but until then you can look here: 2.2.1...2.2.2

2.2.1

PR #1871 Disable setting custom parameters of None or boolean type in web UI. Fixes #1870

2.2.0

• Display locustfile and tasks ratio information on index.html
• Add --autostart and --autoquit parameters (#1864)
• Add constant_throughput wait time (the inverse of constant_pacing)
• Alternative way to rename requests (particularly useful when using an SDK that wraps requests) (#1858)
• Add --equal-weights flag (#1842)
• HttpUser: Unpack known exceptions
• Various charting fixes
• Add FastHttpUser directly under locust package
• Auto-generate Locust's version number using setuptools_scm and git tags
• Show custom arguments in web ui and forward them to worker (#1841)

2.1.0

• Fix docker builds (2.0 final never got pushed to Docker Hub)
• Bump dependency on pyzmq to fix out of memory issue on Windows
• Use 1 as default for user count and spawn rate in web UI start form
• Various documentation updates

2.0.0

Time for a new major version!

This release contains changes that could break/change the behaviour of some existing locust files (but probably not many).

User ramp up/down and User type selection is now controlled by the master instead of autonomously by the workers

This has allowed us to fix some issues with incorrect/skewed User type selection and undesired stepping of ramp up. The issues were especially visible when running many workers and/or using LoadShape:s. This change also allows redistribution of Users if a worker disconnects during a test. This is a major change internally in Locust so please let us know if you encounter any problems (particularly regarding ramp up pace, User distribution, CPU usage on master, etc)

Other potentially breaking API changes

• Change the default User weight to 1 instead of 10 (the old default made no sense)
• Fire test_start and test_stop events on workers too (previously they were only fired on master/standalone instances)
• Workers now send their version number to master. Master will warn about version differences, and pre 2.0-versions will not be allowed to connect at all (because they would not work anyway)
• Update Flask dependency to 2.0

Significant merged PR:s (and prerelease version they were introduced in)

• Allow workers to bypass version check by sending -1 as version (2.0.0) #1830
• Improve logging messages and clean up code after dispatch refactoring (2.0.0b4) #1826
• Remove user_classes_count from heartbeat payload (2.0.0b4) #1825
• Add option to set concurrency of FastHttpUser/Session (2.0.0b3) #1812
• Fire test_start and test_stop events on worker nodes (2.0.0b3) #1777
• Auto shrink request stats table to fit terminal (2.0.0b2) #1811
• Refactoring of the dispatch logic to improve performance (2.0.0b2) #1809
• Check version of workers when they connect. Warn if there is a mismatch, refuse 1.x workers to connect (2.0.0b1) #1805
• Change the default User weight to 1 instead of 10 (2.0.0b1) #1803
• Upgrade to Flask 2 (2.0.0b1) #1764
• Move User selection responsibility from worker to master in order to fix unbalanced distribution of users and uneven ramp-up (2.0.0b0) #1621