Pewpew

Pewpew is an HTTP load test tool designed for ease of use and high performance. See the guide for details on its use.

Changelog

v0.5.6

Changes:

• Add headers_all property to both request and response to allow the access of multiple header values which share the same header name.
• Change the format of the stats output file so data can be appended to it throughout a test run and less data has to be kept in memory.
• Tweak auto sized buffers to grow anytime there is a "endpoing was delayed waiting for provider" event.

Bugfix:

• Fix regression introduced in v0.5.5 where specifying a provider multiple times in a provides would only use the last specified one.
• Fix bug where non-ascii characters could cause an error reading a file when using the line (default) format.
• Fix bug where the line format of a file provider would incorrectly parse files with lines longer than 8KB.
• Fix issue when, under heavy load, pewpew panics with message "found a left over previous bucket".

v0.5.5

Changes:

• Add an error message into the stats when an endpoint is delayed waiting for a provider.
• Adjust the way auto-sized buffers grow to be less aggressive. Now they have to be filled and then emptied before they will grow.
• Short circuit a for_each when a where does not reference for_each.
• Add a command-line argument, start-at, to start the test at a given time.
• Add optional request_timeout option to endpoints.
• Refactor config parser to provide more helpful messages when an error occurs.

Bugfix:

• Fix regression where certain errors which should be handled during a test were causing the test to crash.
• Fix bug where a load_pattern did not work when doing from: 0% and to: 0%.

v0.5.4

Changes:

• Add stats-file command-line flag to specify the name of the stats file.

Bugfix:

• Fix bug where a logs expression would get null for response and request fields unless they were also references within a provides.
• Fix bug where tests would crash if a response header was a not UTF8 encoded.

v0.5.3

Changes:

• Add a days unit to durations.
• Expand cases when an endpoint can have no peak_load to include the case when the endpoint depends upon a response provider.
• Swap out the JSONPath library for one that supports more JSONPath expressions.

Bugfix:

• Raise an error if for_each is referenced but not defined.
• Fix bug where the config parser would erroneously say there was a recursive for_each.

v0.5.2

Changes:

• Allow an unquoted number to be used when indexing into an object within an expression.
• Change stats to go through stdout instead of stderr.
• Print stats for an endpoint even if it only experienced errors.

Bugfix:

• Fix issue where request.body would display the wrong file name when the request body was a file.
• Fix message displayed as request.body when the request body is a file to be consistent with other similar messages and have double braces as delimiters.
• Fix issue where the config.client.request_timeout and config.client.keepalive were not being parsed properly from the config file.
• Fix issue where an endpoint which provides for a provider it also depends on could cause a try run to hang.

v0.5.1

Changes:

• Tags can now reference providers. Add an additional implicit tag of _id which is the index of the endpoint in the list of endpoints.

v0.5.0

Breaking changes:

• Merge the aggregate stats "connection errors" into "test errors".
• Change the collect and repeat expression functions to make the optional max parameter exclusive.
• For endpoints which have provides change the default send behavior to if_not_full if the endpoint has a peak_load otherwise block.
• When an endpoint has multiple provides which are send: block the blocking will only wait for there to be room in at least one of the providers' buffers.
• Moved the config.general summary_output_format option to be a command-line argument.
• Change the command-line interface to have two sub-commands: run and try. With this change is the ability to select more than one endpoint for a try run and specifying a directory for test results.
• Dropped the alias parameter for an endpoint and renamed stats_id to tags.
• Split static providers out to their own section, vars, and make it so environment variables can only be referenced from within a variable defined in vars. This includes no longer requiring environment variables to be prefaced with a $.
• Rename the static_list provider type to list.
• Allow the same header name to be used multiple times in a request (in compliance with HTTP specs). Headers which are set through config.client.headers can be unset in the endpoints.headers sub-section by setting the value to null.

Bugfix:

• Fix an issue where loggers would not log for a request if a warning level error happened.
• Fix performance regression introduced in v0.4.9.
• Fix performance regression introduced in v0.4.10.
• Fix bug where auto_buffer_start_size was not working for response providers.
• Fix bug where some json outputs would have \n at the end.

Changes:

• Allow a logger to log test errors. This is done by making another variable available to loggers--error (in addition to request, response and stats). A logger can no longer have an error "indexing into json", instead it will resolve to null. This enables a logger which logs test errors along with the request and response if they are available.
• Add the watch command-line flag for the run subcommand, along with the watch_transition_time general config option to allow load_patterns and peak_loads to change while a test is running. This enables the duration of a test to change and the amount of load generated.
• Allow a global header to be unset with a null value in the endpoints.headers sub-section.
• Better handle Ctrl-c killing of the test and persist any unsaved data to disk.
• Add in replace expression function which replaces any occurances of a string in a JSON value with another string.
• Add in base64 option for the encode function.
• For response bodies which cannot first be parsed as utf8 strings the response body will be set to the string "<<binary data>>".
• Add in file and format command-line options to try subcommand.
• For multipart requests include the content-length header instead of sending a chunked request.
• For multipart requests which have file pieces, include the filename section with the content-disposition sub-header.
• Allow the selecting of bodies from multipart or file requests. Sections coming from a file are replaced with the string "<<contents of file: "filename">>".

v0.4.11

Changes:

• Print a summary at the end of a try run.
• Allow durations, percents and peak_load to be templates which can interpolate environment variables.
• Add an entries expression function which allows iterating over the key/value pairs in an object.
• Add the stem part of the config file to the name of the stats output file.

v0.4.10

Bug fixes:

• Fix issue where endpoints without a peak_load would run infintely if targeted in a try run.
• Enforce that an endpoint without a peak_load must have at least one provides with send: block.
• Fix issue where warnings (like an invalid url), were not being logged during a try run.
• Fix hang up that would happen with on_demand if the "demander" executed before the on_demand endpoint.
• Fix hang up that could happen when a large amount of data was logged.
• Fix issue where requests would be double counted in the stats if there was a warning level error.

Changes:

• Add no_auto_returns endpoint option.
• In a try run include all endpoints which can provide data to a response provider needed by the target endpoint.

v0.4.9

Bug fixes:

• Fix issue with on_demand where things would hang if the endpoint did not provide a value.
• Fix issue where a logger causes a fatal test error if writing to the underlying resource causes a blocking call.

Changes:

• Add in random and repeat parameters for static_list providers.
• Persists aggregate stats to disk every time they are printed to the console.
• Add in ability to have multipart request bodies.

v0.4.8

Bug fixes:

• Fixed error where a try run could try to use an endpoint which provides for a provider that it also depends on.
• Prevent errors from arising due to boolean operators not being evaluated lazily.
• Fixed error where logger limits only worked when used in conjunction with kill.

Changes:

• Add modulus % expression operator.
• Add repeat option for range providers.
• Disabled load_patterns during try runs.
• Have the process return an error code on a fatal error.
• Add console message when test ends early due to a provider ending.
• Add random expression function.
• Change RTTs to be stored with microsecond precision but still display as milliseconds (with decimal).
• Make some errors which may occur during a test run just print warnings.
• Add log_provider_stats config option to log statistics about providers.
• Add in ability to specify a file as the request body.
• Add on_demand endpoint option.

v0.4.7

Bug fixes:

• Fixed memory leak introduced in v0.4.6.
• Fixed error where order of operations were not being executed correctly--introduced in v0.4.6.
• Fixed off-by-one bug for loggers killing a test.
• Fixed bug where loggers would not log to stdout or stderr.

v0.4.6

Bug fixes:

• Fixed bug where auto_returns were getting blocked waiting for provides to finish.

Changes:

• Add max_parallel_requests endpoint parameter.
• Add the ability to reference environment variables using templates within a static provider.
• Add basic math operators to expressions: +, -, *, /.

v0.4.5:

Changes:

• Refactored the way the code handles errors on the backend. The end result for users is cleaner error messages when things go wrong.
• Added in a "try run" option to the command line. This gives the ability to run one endpoint a single time. Based on the provider needs of the endpoint pewpew will automatically pick any other endpoints needed as dependencies to provide the data. Currently raw request output is automatically printed to stdout.
• Added in a three argument version of the join expression function which enables the string formatting of an object.

v0.4.4:

Changes:

• Add in proper handling of compressed response bodies. Brotli, gzip and deflate are supported.

v0.4.3:

Changes:

• Add range function which creates an array of numbers. See docs for details.
• Added min and max functions which select the smallest or largest numbers respectively in a series of numbers.

v0.4.2:

Bug fixes:

• Fixed bug where when a request failed auto_return values were not being sent back to their provider.

Changes:

• stats_id can now have templates as the values.
• Previously as a test progressed, the url printed in the stats output could change because dynamic parts of the url were replaced by a star (*). Now we detect the dynamic parts of a url at the beginning of a test so it should start out with the stars.

v0.4.1:

Bug fixes:

• Fixed bug where variables in declare were not being recognized as providers in an expression.
• Fixed bug where bucket stats were sometimes printed in the middle of test summary stats.

v0.4.0:

Breaking changes:

• Templates no longer use {{ }} and now use ${ }.
• All helper functions use the more familiar parenthesis as part of the calling syntax and commas to separate arguments. Ex: what was {{join foo "-"}} is now ${join(foo, "-")}.
• There is no longer an environment provider. Instead environment variables can be referenced in any expression using $ENV_VAR syntax. If you previously referenced an environment variable in a file path, make sure you change it to have the $ before the variable name.
• The declare section now supports any valid expression.
• The order of collect function's arguments has changed.