Remote servers The Goal: To scale stress testing up by allowing testers to add servers and control them remotely through a single gui instance that integrates the results from all the servers.
The Reality: 5 remote servers sending every byte from the web server being tested on to the single client gui makes those 5 servers no faster than the single client would have been in any case (because the bottleneck is usually java IO, not client side processing). Further problems: you can't simple create a complex test and go - if there are supporting files, they all have to be carefully placed onto each remote server manually before testing. Ideally, the remote servers should get everything they need from the client machine when you click "go". At the moment, it's pointless. I achieve the same goal by manually running multiple machines in non-gui mode, starting them all at roughly the same time, letting them run for a long time, and then merging the resulting .jtl files into one and loading into a visualizer. Far from ideal, but solving the remote testing issue would take a lot of doing, IMO. -Mike On Thu, 2004-06-10 at 18:00, Remedy QA wrote: > Thanks for the clarification. > So do you recommend us to even use remote testing if you think it's > "ugh"? > I just noticed that global counters are not shared among the servers. > Each server has it's own counter which starts at the same values as the > other servers. That sucks. :-( > > > --- Michael Stover <[EMAIL PROTECTED]> wrote: > > JMeter remote testing - ugh. That's my assessment of the current > > state > > of this functionality. > > > > Anyway, to answer your questions: > > 1. In this case, "client" and "server" are used in a sense such that > > you, the user, sits and does work on the "client", which then sends > > requests to the "servers". Think of an email "client" that polls > > multiple email "server" for messages. The request the client sends > > is > > "do this test", and then the sample results roll in. > > > > 2. The client sends the whole test to every server. So, if the test > > specifies 100 threads, each server runs 100 threads. I think you can > > figure out the degree of control you have given that... > > > > 3. I don't know anything about the hold_samples property. > > > > 4. Right. > > > > -Mike > > > > On Thu, 2004-06-10 at 15:19, Remedy QA wrote: > > > I am confused as to the setup of distributed testing. I have read > > the > > > docs on how to set up but the use of server vs. client seems > > backwards > > > to me. I got the connections going but I'm not really > > understanding > > > how it works. > > > > > > According to the docs, or how I am reading into them, there are > > many > > > JMeters running in server mode, therefore, many RMI registries > > running > > > on separate machines. For example, S1, S2, S3 are machines running > > > Jmeter server. I am thinking there is no GUI or batch jmeter > > running > > > at all, just the jmeter server. > > > > > > Then according to the docs, there is only one client controller, > > which > > > uses the GUI to control. I'll call this machine C1. In C1's jmeter > > > properties file, it has S1, S2, and S3 listed as the remote_hosts > > > values. Then if C1's test plan has 100 thread users and I select > > > Run/RemoteStartAll, it will start the remote testing on all the > > server > > > machines. > > > > > > The parts where I am confused: > > > 1. The use of server and client is backwards to me. Shouldn't > > there > > > only be one server and many clients? And the server should be the > > > controller. > > > > > > 2. How on C1, if I specified 100 thread users, does it distribute > > > among the 3 server machines? Is it always divided evenly? Can I > > > control the distribution? > > > > > > 3. Are the results gathering done while the load test is running? > > I > > > read in previous archived messages that there is a "hold_samples" > > > property that you can set to indicate all results should be written > > at > > > the end of the test. However, I don't see such a property setting > > in > > > jmeter.properties. Then is the gathering of results done on C1? > > > > > > 4. Since C1 is the controller, it is not sending HTTP Requests, > > only > > > facilitating and from #3, gathering results? > > > > > > Thanks for the clarification! > > > mabel > > > > > > __________________________________ > Do you Yahoo!? > Friends. Fun. Try the all-new Yahoo! Messenger. > http://messenger.yahoo.com/ > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] -- Michael Stover <[EMAIL PROTECTED]> Apache Software Foundation --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
