On Sun, 18 Jan 2004, Moshe Matitya wrote:
> I am trying to figure out how to tell Polygraph to use unlimited
> persistent connections, but have become a bit confused.
>
> The documentation for the 'Agent' type
> [<http://www.web-polygraph.org/test/docs/reference/pgl/types.html#type:docs/reference/pgl/types/Agent>]
> seems self-contradictory:
>
> "The 'http_versions' selector ... has no effect on other defaults.
> For example, you still need to explicitly enable persistent
> connections, even if you are using HTTP/1.1 agents." This seems to
> imply that the default is *non-persistent* connections.
Correct.
> But a couple of sentences later it says: "An HTTP connection will
> never have more than 'pconn_use_lmt requests'. To disable
> persistent connections, set use limit to 'const(1)'." This seems to
> imply that the default is *persistent* connections, after all.
Please see if the current wording is any better. It now explicitly
documents what the default is.
> Which one is it? More to the point, how do I specify that I want
> *unlimited* persistent connections?
It is documented now. Please try the following:
pconn_use_lmt = const(2147483647);
> Surely I don't have to use some arbitrarily "very big" number?
> This would problematic (in addition to ugly), in the event that the
> number of uses should actually approach the "very big" number.
Sorry to disappoint you. I agree that it is ugly, but PGL does not
have a way of expressing "unlimited" or "infinite" values yet. You can
make it a little bit nicer using the following trick:
int Infinity = 2147483647;
pconn_use_lmt = const(Infinity);
Note that you are unlikely to approach that limit because of
real-world constraints on response time: even with 1ms response times,
it would take 24 days to submit that many requests. Moreover, I
suspect that some internal Polygraph counters will break if you do
manage to submit that many. Polygraph can be used to run weeks-long
tests, but we have not tested it at MAX_INT scale.
> I'm sure there must be some plain, obvious way to do this.
Yes (see the above). It is plain and obvious, but ugly :-). I have not
tested the above, so please let me know if it does not work as
intended.
HTH,
Alex.
-- Protocol performance, functionality, and reliability testing. Tools, services, and know-how. http://www.measurement-factory.com/
This archive was generated by hypermail 2b29 : Mon Feb 06 2006 - 12:00:27 MST