Alex,
I have the same problem that polysrv quits and polyclt just keeps
going on and on in the 'cool' phase. This happens in spite of the fact that
I have specified a 5min idle_tout for polysrv.
I was running polymix-1, with just one server and one robot. And, I
noticed that the reply rate is a lot higher on the polyclt than on the
polysrv side. Could this be a cause of the above problem?
polysrv::
006.28| i-meas 8587 22.99 3048 0.00 0 65
006.36| i-meas 8687 20.00 3080 0.00 0 83
006.44| i-meas 8814 25.39 2992 0.00 0 68
006.53| i-meas 8927 22.60 3051 0.00 0 61
006.61| i-meas 9028 20.19 3049 0.00 0 80
006.69| i-meas 9162 26.78 3007 0.00 0 76
006.78| i-meas 9282 24.00 2875 0.00 0 83
006.86| i-meas 9405 24.60 3294 0.00 0 70
polyclt::
006.20| i-meas 17589 54.20 17381 56.46 0 1038
006.28| i-meas 17825 47.18 22008 56.78 0 1065
006.37| i-meas 18085 51.99 21129 51.54 0 1069
006.45| i-meas 18326 48.20 15789 55.60 0 1054
006.53| i-meas 18581 50.99 17244 51.76 0 1055
006.62| i-meas 18837 51.19 19803 55.08 0 1094
006.70| i-meas 19101 52.80 20702 55.68 0 1075
006.78| i-meas 19355 50.78 13061 56.69 0 1095
Thanks,
Sadhna.
> -----Original Message-----
> From: ext Alex Rousskov [mailto:rousskov@measurement-factory.com]
> Sent: Friday,April 13,2001 8:19 PM
> To: Lynn Scott
> Cc: polygraph@ircache.net
> Subject: Re: polyclt shutdown
>
>
> On Fri, 13 Apr 2001, Duane Wessels wrote:
>
> > On Fri, 13 Apr 2001, Lynn Scott wrote:
> >
> > > Dear Sirs,
> > >
> > > I previous inquired of Alex on using the the --notify
> option to signal
> > > my process when Web Polygraph was complete. Alex
> mentioned that if
> > > properly configured, Polygraph processes quit when the
> test is over. In
> > > doing some testing, I am running polymix-1 on a single
> workstation with
> > > a very small duration expirement.
> > >
> > > // experiment duration
> > > time ExpDur = 1min
> > >
> > > After the tests complete the server process (polysrv)
> shutsdown properly
> > >
> > > 002.22| got 698 xactions and 0 errors
> > > 002.22| shutdown reason: all 4 phases met their goals
> > >
> > > However, the client process continues to run.
> > >
> > > 002.35| i-cool 870 3.79 45001 100.00 18 28
> > > 002.43| i-cool 870 0.20 -1 -1.00 1 27
> > > 002.52| i-cool 870 3.19 -1 -1.00 16 11
> > > 002.60| i-cool 870 2.00 -1 -1.00 10 1
> > > 002.69| i-cool 870 0.00 -1 -1.00 0 1
> > > 002.77| i-cool 870 0.00 -1 -1.00 0 1
> > > 002.85| i-cool 870 0.20 -1 -1.00 1 0
> > >
> > > How do I properly configure so that both will shutdown?
> >
> > You might have to wait just a bit longer. I think the cool phase
> > is hard-wired at one minute. Also the cool phase might not exit
> > as long as there are outstanding requests (the last column).
>
> Not exactly. There are no hard-wired phases with non-zero minimum
> duration in Polygraph. When all _configured_ phases are over and no
> client-side idle timeout is specified, polyclt should exit. Polygraph
> does have a dummy "cold" phase that starts right after all configured
> phases are done, but that phase has no minimum duration and is usually
> not even visible.
>
> My guess is that you are using Polygraph version 2.5.4 and you did not
> specify idle timeout for the server side. Your servers quit
> before clients. Recent polygraph versions synchronize phase schedules,
> and when polysrv quits first, that synchronization breaks -- clients
> wait for servers to confirm phase change for ever.
>
> The solution is to use "--idle_tout 3min" option (3min can be changed
> to something smaller if needed) with polysrv and polysrv only. This
> will be the default starting with the next release.
>
> HTH,
>
> Alex.
>
>
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:18 MDT