SleepGenerator Guide

SleepGenerator is a client that generates (usually many) calls of a SleepingEcho Web Service (a service that receives a message and after a period of sleep it returns it back - really an exciting service; its code is in org.bionanny.samples.SleepingEchoImpl).

The main purpose is to have enough calls that can be monitored, even the calls that can be monitored in real-time (to see so-called still-running requests). It also tests your Tomcat's configuration because it produces a *lot* of threads - so usually the Tomcat complains first but should still respond to all calls, eventually (except that some calls are cancelled because they time-out).

It makes sense to use this client only after you have deployed one or more SleepingEcho Web Services, and others. All that can be done by an Ant target deploy - see the build guide.

	Usage

The SleepGenerator is started by:

build/run/run-generator [options] -f <property-file>
build/run/run-generator [options] <property-file>

This actually calls the class org.bionanny.samples.SleepGenerator.

	Options

Use option -help to get (perhaps more accurate) list of all options.

-f <property-file>	This is the main - and mandatory - option (the leading -f may be omitted). The file format is described below.
-v	verbose mode (it reports every call)
-q	more quiet mode
-c	With this option, the client just checks the format of the input property file and prints what would be called, but does not do any real calling (but it still simulates all calls for the full time as indicated in the property file; so it can run long).

	Sleeping properties file

The most important command-line parameter is the <property-file>. It defines what services to call (all of them have always the same functionality, they just sleep for a while, but they are differently named as they constitute several/many Web Services).

The file has format of a Java property file, with the following properties and rules:

• The main property (used usually more than once) has one of the following formats:

  sleeping.call[.<number>].<service-name> = [<how-long-to-sleep-in-millis>]
  sleeping.acall[.<number>].<service-name> = [<how-long-to-sleep-in-millis>]
         

• Each property sleeping.call represents an invocation of a web service given in <service-name>. Also each property sleeping.acall does that - but this time for an asynchronous calls (see Java interface org.bionanny.samples.SleepingEcho what is the difference between synchronous and asynchronous calls).
• <service-name> is either a full service endpoint, or the last part of an endpoint (just the pure service name): if it can be converted to a valid URL, it is considered as a full endpoint, otherwise it is prefixed by the value of a property sleeping.endpoint.
• <number> is actually ignored - it is here only to allow defining more calls to the same service without necessity to use the same property name more than once (which would cause implementation problems).
• Each service invocation will get as a parameter <how-long-to-sleep-in-millis> (which indicates how long the service will sleep before returning back); if it is not set, a random numer of milli-seconds will be generated; the range of this random number is <10, 3 * 60 * 1000>; (up to three minutes) unless properties sleeping.min and/or sleeping.max is/are given.
• All service calls will be repeated just once unless property sleeping.repeat is given. Its value indicates how many minutes to repeat all defined calls. If its value is empty or negative, the repetition is infinitive.
• All service calls will be done in the (numerically-sorted) order of all call definitions (so this is yet another purpose for the <number> in the call and acall property names) unless property sleeping.shuffle with an empty or has true value.
• Before any call is made a property sleeping.maxdelay is checked. If it does not exist, or if it contains a positive number, then the call is delayed for a random number of milliseconds in the range <0, sleeping.maxdelay>. This property makes life easier for the Tomcat - so if you wish a full-stress test, define it with a zero value.
• Summary:

  Name	Default value
  sleeping.endpoint = <endpoint>	http://localhost:8080/axis/services
  sleeping.min = [<lower-limit-in-millis>]	10 ms
  sleeping.max = [<upper-limit-in-millis>]	3*60*1000 ms
  sleeping.repeat = [<how-many-minutes-to-keep-calling>]	0 mins
  sleeping.shuffle = [true | yes | 1]	true
  sleeping.verbose = [1 | true | yes | 0 | false | no]	false
  sleeping.maxdelay = <delay-efore-a-call-is-made>	5*1000 ms

There is a testing property file (created by Ant during the compilation time) that can be used (directly or as a template): build/data/sleepers.input. That's why, the sleep generator is usually invoked by typing:

build/run/run-generator build/data/sleepers.input

Probably the best is to use the SleepGenerator together with a client that can show on-the-fly collected data. Such client is a cgi-bin Perl client.