On Apr 10, 5:38 pm, Thomas Krüger <[EMAIL PROTECTED]> wrote: > Godzilla schrieb: > > > > > > > I have been using the queue module for a multithreaded environment and > > things seem to work well... until we had a requirement for the > > application to be able to time sync to the server. With the time sync, > > it actually disorientated the timeout in the queue's get() method... > > e.g. > > > get(item, 2.0) > > > After the time sync, say 15 seconds backward, the thread is sitting on > > that get() method for a total of 17 seconds. We can only sync the > > device once per day and the time can drift up to 15 seconds per day!! > > I had tried to get around this problem by having a sleep(2) (sleep is > > not local system time dependant) just before the get(), but that will > > slow down the application too much. > > > Anyone knows a solution to this problem or an alternative method? > > I was fixing a serious time drift problem on Linux lately. If your > server runs on Linux I can give you some hints: > > - set the system clock > - delete /etc/adjtime and resync system time and hardware time > "hwclock --systohc" > - on some distributions you may have to reboot > - give it some time to see if it is still drifting > - if only the system clocks drifts (see output of "hwclock; date") > you may have a timer related kernel problem. This may help: > * deactivate ACPI via kernel boot parameter > * change timer frequency > * try different setting for all timer related stuff like > CONFIG_HPET_TIMER or CONFIG_X86_PM_TIMER > > Thomas > > -- > sinature:http://nospam.nowire.org/signature_usenet.png- Hide quoted text - > > - Show quoted text -
Thanks Thomas, I'm not running Linux but I will take some of your pointers as a guide. Cheers mate. -- http://mail.python.org/mailman/listinfo/python-list
