There was a bug fix for the problem you are describing in 2.2.5.
If you are not using 2.2.5, you should. If you are suing 2.2.5 please
forward a binary log (with several phases that does not get processed by
lx) to firstname.lastname@example.org.
If you do not specify --phase, lx merges data from all phases.
P.S. I assume that console output (or lr) shows that the phases you are
looking for have actually happened.
On Thu, 2 Dec 1999, John Dilley wrote:
> I am trying to use polygraph to characterize the performance of some of our
> Akamai servers. I've gotten two polyclt systems and a polysrv to generate
> sufficient load to saturate a server. Now I want to examine the results. I
> have a PGL file with a couple of phases, but lx --phase does not print any
> useful information (all the results come out either -1 or 0). If I say
> lx --objects All without a phase I get a histogram, which appears to be
> valid. I can not tell if it is for the first logged phase (the first one
> with log_stats=true), the second, or both.
> What I really want is to be able to have a set of phases, each at a specific
> load level (req_rate), and look at each of them separately. I could do this
> by creating 5 or so PGL files, but then I have to set up the server and
> poly* 5 times; I'd prefer to avoid this if possible. What do you guys use to
> get response time by offered load in the bakeoff runs? Am I missing
> something obvious? Thanks for the help,
> -- jad --
> John Dilley
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:10 MDT