jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Shmuel Krakower <>
Subject Re: Issues 52673 / 52698 / 41545 / 52707 / 52502
Date Tue, 20 Mar 2012 09:14:48 GMT
Hi Geoff,
I find this capability really cool: allow to load test a system with the
exact same real life usage.

I find the most problamatic issue with load testing is that you cannot
almost never run a load test which is simulating by 100% the real life
usage by covering all of the featured being used on the SUT.
Thus people can invest on load testing some part of the system and I almost
never create a load test script which really covers all the features. This
is causing me some times to say "ok the system is ready to
go-live(according to the limitation of the load testing scenaruo)" and then
we find that some users are actually running some heavy tasks which were
not part of the load test scenario, which kills the system.

So having a way to cover with a load test the exact real life usage is a

The thing is that I think that this sampler won't really do that. I am not
sure that this may really work for me for even one of load testing project
I've working one in the last few years. The main issue I have with this is
that running the same requests as recorded is just cannot work. There is
almost no system that doesn't use cookies, sessions and other dynamic
content which just cannot simply be replayed.

It might be that your sampler does this and if it is, it may work (I didn't
look into details). Can you share your use of it and how you overcome these

Shmuel Krakower.
בתאריך 2012 3 20 10:35, מאת "Geoff Simmons" <>:

> Hash: SHA256
> On 3/19/12 11:23 PM, sebb wrote:
> > On 19 March 2012 20:31, Philippe Mouawad
> > <> wrote:
> >>
> >> Shall we integrate this sampler:
> >>
> >> -
> >>
> >> My opinion regarding it is that as it's difficult to generate
> >> data usable by this sampler using it will be hard.
> >>
> >> So I don't have a definite opinion.
> >
> > If it is implemented, why not implement it as an Access Log
> > Sampler?
> >
> > But I agree - I'm not sure it's worth it.
> Hey all, if I may comment on this -- if the information in an access
> log is sufficient for the needs of a test, then the Access Log Sampler
> is a great deal easier to use.
> The record/replay scenario is indicated for a test that depends on any of:
> - - contents of request bodies, such as POST data
> - - contents of request headers that aren't available from access logs
> - - reproducing the concurrency of the recorded traffic
> For that, you need logs of the full contents of requests, either
> recorded or synthetic. And you're right, that's quite a bit different
> from just dropping in an access log.
> So the Replay Sampler certainly wouldn't be used by everyone -- you'd
> have to be willing and able to do the preparations, for a test that
> needs it. Whether that's of sufficient interest for the JMeter
> distribution is a judgment call, but you're welcome to it (since a
> number of JMeter components are fairly specialized, it didn't seem out
> of the question). I'll be continuing to maintain it, and would do so
> in the context of the project.
> Best,
> Geoff
> - --
> UPLEX Systemoptimierung
> Schwanenwik 24
> 22087 Hamburg
> Mob: +49-176-63690917
> Version: GnuPG/MacGPG2 v2.0.14 (Darwin)
> Comment: Using GnuPG with Mozilla -
> LRO7nPa8PoRHyePYvALAQkoiEtqpiwzDUHgonVHHJeSkLJgArC8xPtEt2JgY5VAA
> keFLkJlL9oUF+DTVAa5fAEv0r3ThFK1LUebKkc2tQJR/0Rv/aAQ3W/tU447pppKU
> O5m7rC1p2zdA4DUZgEmRwInamD4WoEc7WNRCQDvsvWX8FdezvezGalWsrTKFr5gf
> zkmdTS89RpEOY5rpolsKxHyoJA0zmZD5pyHUaItYdUTQbAmZ8unCgM7WNAslClMw
> lQpaOaAcuwTTd6e45YC6tQcXBQtqfqY2T1b6+SPvFEJ2W1PojWwBGmufBb5x0wfR
> qN2Z4BgxpTsdtluxigshMmoS37XqGZR0RX98/fLkec80kqA8590vMwT0zIRaw1qE
> wQp8R0tUKXoYRwwr2kkg5MSW5blCfUgiGnJjcE0iPi2TRicXcg2H4mFVTdElf/R5
> cf/YPDGw2wAfrdVqreeygHQJqd/omX6ui1kKqTJQl6F4RyoAwxJcc6aDxH5HI6sh
> kgC9PXlzVVfE49XyuXYIHxMv0jsxfdLfbbkZzgd64EY8H1TyOIXx0A27G07WgiHa
> QGINBLhI+DJvPiG3NrJE6+Xq43k9lPX4boQna6ifOrLIrKAcTxeJGypxnptnpSMV
> cx8+ac0l1C7h8aAIh3I3
> =jDcM

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message