cjimison/barrage

Scalable http load testing framework

barrage's Issues

Add Support for Filter nodes

Add support for random selection node

The random selection node should pick one node from all its children and execute that node. You can also set a "count" value as to how many times you want to randomly select the node.

It would also be cool if I added a field for "timed" wait. The idea would be you will randomly select a block on child nodes for up to N amount out time. This could simulate a game sessions or something alongs those lines.

Cookie support is not working AGAIN!!!

Looks like when I move the inets:httpc to use profile based commands it broke cookies. I need to fix that.

Make is so the general sends the actions to the commanders before they execute.

The issue here is I want commanders to not have to know the actions they are executing. This way in theory we can just a have a bank of commanders running and the users will just start up their own general and tap into any commanders waiting around.

Results report

Now that I am running stuff, it would be great if I could get an actual report with the timing of the requests. This data should be able to feed into a graph. JSON maybe a good format given that I think I will just stick a front end over the load testing system in HTML

Add support for https requests

Not sure how important this is right now. But is would be great if I had support for https as well as http.

Allow Array access and data cache object movements

I want to be able to manipulate the data returned from the load test in the tree. Simple things.

Buttons appear in reverse order

Buttons at the top of the page appear in reverse order as defined in behaviors.json

Distributed commanders are broken AGAIN!

Looks like the general is not sending off the attack requests to all the commanders except for the one it creates assuming it creates one. I need to figure out what went wrong here!

The the ability to change the name of the default files to load

I want to be able to change the name of the data.json files so I can switch behavior sets without having to back up the file and do a bunch of crap like that.

Don't use sans serif font

Suggestion on how to make the webpage look a little nicer :)

Periodic behaviours

Have a way to define behaviours that run every X seconds. For example: check inbox every 30 seconds.

Input data source for actions

Have a data source to choose certain request parameters. Let's say we could have a json file to choose random users at the login action. Or have another json to pick quest ids sequentially.

Support for JSON results

99.99% of server return JSON blobs these days. I should add some support for that...

composite behaviors

It would be super cool if you can add a "behavior" as a child to a node. That way people can make stuff like "login" that may take 3 or for "actions" and then just reference that.

Add support for JSON based config files

Right now all my config files are based off of erlang. It would be better if these were JSON based so I can show them off better.

Refactor barrage_commander to deal with fault better

Right now if a gunner gets stalled out or put in a bad state the whole commander is in lockdown. I really need to handle fault better in the commander given that this system is testing other systems so error should be expected.

Streaming data results

On thing that really sucks at the moment is you don't get to see any results or know the status of you test until it is completed. Not a big problem if you test is short, but if it is long running this can be a very big problem. I need to refactor the system (and the web interface) to allow for streaming input data.

Add Support for cookies

A lot of system uses cookies. I just need to support it.

Allow to reference json attributes in the key value cache store

In order to allow to do something like this: $USER.id

Add support for conditional nodes

I should add conditional support node

General does not wait for all commanders to report back data before sending results

Currently when the general gets the first message to "report_results" from a commander, it will assume that all testing is done. This will not work when there are multiple commanders.

Add KVS Index value for a count based behavior

This way the script can know how many times this has executed.

This should be a named param given from the user. If the name exists then the data is stored there.

Add http code response stats

Add a graph or table to show a breakdown of responses grouped by http code.

Make the General page port configurable via the barrage.json file

It was pointed out that I should add the ability to configure what port I server up http content for the general controller page. Shouldn't be that hard.

General will issue more orders out when it gets a request.

When the barrage_general gets a call for "issue_order" is should validate that it is not currently running an order.

Better sorting of results

Currently with the block run the sorted results sucks. I should log the order data is received by the gunners and sort based on that. This would give a better "time" prediction on what the response times are.

Add License to source files

I need to add the license agreement to all the source files.

Show more detailed stats per action

Apart from the graph per action, show a summary with more useful stats like: average, min, max , std dev.

Refactor barrage_gunner to deal with fault

Currently the system is very my way or the highway kind of thing. The only problem is this system is testing other systems that maybe buggy. I need to verify all incoming data (such as actions definitions, or JSON from servers), and if it fails, report it as a failure for the test. Not just fail the process which will hang the commanders and stall the generals.

Add HTML front end to General so I can launch attacks and check on system health

I would like three things to start with.

1. To launch an attack from the webpage
2. To see what Commanders are available (and now many gunners their are)
3. Results of the attack

Get rebar release build working

The release build is not getting generated correctly. I need to get that setup up again!

Negative values in X-Axis popup

I stumbled upon several of these popups after running tests for a while ( from the streaming page )

Screenshot attached:

Add support for waiting a timed amount in the Behavior tree

This should support both a static and a random wait. In the future I may want to add support for a dynamic wait based on arguments given by the server but I need to think out the use case for that...

Add support for server results storage

For many tests the server will return a value that will be need to be stored and then passed in later as an arg

Support for the barrage.config file

I would like to have support for a basic barrage.config file. This should hold the following data:

1. Type <== Is this a general or a commander node in the cluster
2. Server <== What the load testing server we are barraging
3. Host <== If this is not a general node then who is..

Support "cluster" mode that will allow me to distribute commanders to other machines

I want to support a system kind of like how Riak does it. You startup the main know that will be the "general" then each addition node can be attached to it.

Implement Child processing of behaviors

Currently I am only executing the root node of a behavior and I need to get the child processing working.

Time factor

"wait" and "random_wait" actions could have a time factor either to accelerate or slow down time.

Add support for URLS that have non static data inside of them

Example would be something like
http://localhost/users/<USER_NAME>/deck/<DECK_ID>/cards or what not.

Add support for web based configurations

Right now everything has to be configured via the config files. This is great but it assumes some understanding of how erlang reltools work (vm.config, sys.config). I want to make it possible for people to set these systems up faster without having to understand erlang systems.

replace httpc with ibrowse

From the looks of the graph, my httpc requests are pooling up. I think this is because all httpc calls are being sent to one gen_server that is executing the requests in order. This makes a lot of sense as to why my graphs tend to start off big and then tail down.

Move project to gitflow

Using only master as a branch sucks and is stupid. Git-flow is awesome, is becoming a standard, and has tools around it. So use git-flow!!!

Add Commander monitoring of the general.

Right now if the general goes down, the commander will not know and if a test is running it can get screwed up.

Add ability to have raw header and raw body to request

Add support for configuration Modules

add support for configuration to be broken up into Modules and imported into other Modules, so that complex configuration doesn't have to be one big honking blob of JSON

thanks!

Add the ability to pull non-static "count" values out of the KVS system

The will allow for looping through a bunch of data sent from the server if needed.

Add support for DO action

The do action will execute an action N times. With N being a value passed into the args.

This would be good for preparing a DB with random data

USECASE : OAUTH

I should come up with a use case that tests an oauth based session

Add support for post requests

Currently I am only support post commits. It would be great if I could support more http request types, but hey, lets start simple. Just a post would be nice :)